[{"data":1,"prerenderedAt":2123},["ShallowReactive",2],{"site-nav-content":3,"blog:\u002Fblog\u002Fhow-to-evaluate-an-agent-harness":177,"blog-index-copy":881,"blog:\u002Fblog\u002Fhow-to-evaluate-an-agent-harness:surround":902,"hiring-banner-content":2092,"site-cta-content":2104},{"header":4,"productNav":9,"nav":42,"footer":61,"askAI":132,"id":161,"title":162,"archived":163,"authors":164,"badge":164,"body":165,"date":164,"definedTerm":164,"department":164,"description":169,"extension":172,"eyebrow":164,"faqHeader":164,"faqs":164,"footerBand":164,"headline":164,"image":164,"industry":164,"jobType":164,"listed":131,"location":164,"navigation":131,"openRoles":164,"pageLayout":164,"path":173,"relatedHeading":164,"seo":174,"series":164,"sitemap":163,"status":164,"stem":175,"subhead":164,"tags":164,"video":164,"whyJoin":164,"workplaceType":164,"__hash__":176},{"productLabel":5,"loginLabel":6,"contactLabel":7,"contactSalesLabel":8},"Product","Log in","Contact","Get started for free",[10,14,18,22,26,30,34,38],{"label":11,"to":12,"description":13},"Overview","\u002Foverview","Seven layers. One closed loop.",{"label":15,"to":16,"description":17},"Conflux","\u002Fproduct\u002Fconflux","Where your team, workstreams, and agents meet.",{"label":19,"to":20,"description":21},"Agent Teams","\u002Fproduct\u002Fagent-teams","Specialist teams - governed from day one.",{"label":23,"to":24,"description":25},"Lifecycle Graph","\u002Fproduct\u002Flifecycle-graph","Intelligence that compounds across every interaction.",{"label":27,"to":28,"description":29},"Company Wiki","\u002Fproduct\u002Fwiki","Playbooks and policies where expertise stays.",{"label":31,"to":32,"description":33},"Workstreams","\u002Fproduct\u002Fworkstreams","From brief to signed-off deliverable on one canvas.",{"label":35,"to":36,"description":37},"Perception Console","\u002Fproduct\u002Fperception","Ask your whole business in plain English.",{"label":39,"to":40,"description":41},"Governance","\u002Fproduct\u002Fgovernance","Frontier AI you can actually sign off on.",[43,46,49,52,55,58],{"label":44,"to":45},"Models","\u002Fmodels",{"label":47,"to":48},"Pricing","\u002Fpricing",{"label":50,"to":51},"Integrations","\u002Fintegrations",{"label":53,"to":54},"Security","\u002Fsecurity",{"label":56,"to":57},"Partners","\u002Fpartners",{"label":59,"to":60},"Insights","\u002Fblog",{"productHeading":5,"companyHeading":62,"resourcesHeading":63,"legalHeading":64,"docsLabel":65,"docsUrl":66,"statementLines":67,"copyright":70,"companyLinks":71,"resourcesLinks":86,"legalLinks":102,"socialLinks":109,"bottomLinks":119},"Company","Resources","Legal","Docs","https:\u002F\u002Fdocs.gonimbus.ai",[68,69],"Stop training someone else's model.","Control your AI.","© 2026 Nimbus Intelligence, Inc. All rights reserved.",[72,73,74,75,76,78,81,84],{"label":47,"to":48},{"label":50,"to":51},{"label":53,"to":54},{"label":59,"to":60},{"label":77,"to":57},"Partner Program",{"label":79,"to":80},"Careers","\u002Fcareers",{"label":82,"to":83},"System status","\u002Fstatus",{"label":7,"to":85},"\u002Fcontact",[87,90,93,96,99],{"label":88,"to":89},"Glossary","\u002Fglossary",{"label":91,"to":92},"Compare","\u002Fcompare",{"label":94,"to":95},"Evaluate","\u002Fevaluate",{"label":97,"to":98},"Problems","\u002Fproblems",{"label":100,"to":101},"Use cases","\u002Fuse-cases",[103,106],{"label":104,"to":105},"Terms of Service","\u002Fterms",{"label":107,"to":108},"Privacy Policy","\u002Fprivacy",[110,113,116],{"label":111,"href":112},"LinkedIn","https:\u002F\u002Fwww.linkedin.com\u002Fcompany\u002Fgonimbusai\u002F",{"label":114,"href":115},"X","https:\u002F\u002Fx.com\u002Fgonimbusai",{"label":117,"href":118},"Instagram","https:\u002F\u002Fwww.instagram.com\u002Fgonimbus_ai\u002F",[120,122,124,127,128],{"label":121,"to":105},"Terms",{"label":123,"to":108},"Privacy",{"label":125,"to":126},"Compliance","\u002Fcompliance",{"label":82,"to":83},{"label":129,"to":130,"external":131},"LLMs.txt","\u002Fllms.txt",true,{"text":133,"prompt":134,"platforms":135},"Ask AI about Nimbus","I'm researching enterprise intelligence platforms and want to know how Nimbus combines perception, collaboration, and autonomous agents to drive strategic decision-making. Summarize the highlights from Nimbus's website: https:\u002F\u002Fgonimbus.ai",[136,141,146,151,156],{"name":137,"label":138,"icon":139,"hrefPrefix":140},"chatgpt","ChatGPT","simple-icons:openai","https:\u002F\u002Fchatgpt.com\u002F?prompt=",{"name":142,"label":143,"icon":144,"hrefPrefix":145},"perplexity","Perplexity","mdi:magnify","https:\u002F\u002Fwww.perplexity.ai\u002Fsearch\u002Fnew?q=",{"name":147,"label":148,"icon":149,"hrefPrefix":150},"grok","Grok","simple-icons:x","https:\u002F\u002Fx.com\u002Fi\u002Fgrok?text=",{"name":152,"label":153,"icon":154,"hrefPrefix":155},"claude","Claude","simple-icons:anthropic","https:\u002F\u002Fclaude.ai\u002Fnew?q=",{"name":157,"label":158,"icon":159,"hrefPrefix":160},"google-ai","Google AI","simple-icons:google","https:\u002F\u002Fwww.google.com\u002Fsearch?udm=50&aep=11&q=","content\u002Fshared\u002Fnav.md","Site navigation",false,null,{"type":166,"value":167,"toc":168},"minimark",[],{"title":169,"searchDepth":170,"depth":170,"links":171},"",2,[],"md","\u002Fshared\u002Fnav",{"title":162,"description":169},"shared\u002Fnav","Q7sDe7TuGEwhwUMamyeOxjvGlsUaF3Iay9VTW0KaE_U",{"id":178,"title":179,"archived":163,"authors":180,"badge":184,"body":186,"date":858,"definedTerm":164,"department":164,"description":859,"extension":172,"eyebrow":164,"faqHeader":860,"faqs":863,"footerBand":164,"headline":164,"image":164,"industry":164,"jobType":164,"listed":131,"location":164,"navigation":131,"openRoles":164,"pageLayout":164,"path":873,"relatedHeading":164,"seo":874,"series":875,"sitemap":131,"status":164,"stem":876,"subhead":164,"tags":877,"video":164,"whyJoin":164,"workplaceType":164,"__hash__":880},"content\u002Fblog\u002Fhow-to-evaluate-an-agent-harness.md","How to Evaluate an Agent Harness",[181],{"name":182,"to":183},"Nimbus Research","https:\u002F\u002Fgonimbus.ai",{"label":185},"Evaluation",{"type":166,"value":187,"toc":838},[188,198,220,237,242,315,328,332,335,369,376,379,383,398,415,433,443,458,469,484,495,506,511,537,540,544,550,560,563,567,570,579,585,596,606,615,623,631,643,650,653,660,664,690,694,698,701,705,710,714,717,721,724,728,743,747,757,761],[189,190,191,192,197],"p",{},"Evaluating an ",[193,194,196],"a",{"href":195},"\u002Fblog\u002Fwhat-is-an-agent-harness\u002F","agent harness"," is checking whether the runtime around the model can finish a job under a stop you trust — not whether a demo answered a question.",[189,199,200,204,205,209,210,214,215,219],{},[193,201,203],{"href":202},"https:\u002F\u002Fdocs.langchain.com\u002Foss\u002Fpython\u002Flangchain\u002Fagents","LangChain"," defines the object: Agent = Model + Harness. The scoring sheet is therefore about the harness. If your RFP starts with context-window size and SWE-bench, you are scoring a model (and maybe an inner coding loop). You will miss whether an unsigned Salesforce PATCH is possible. ",[193,206,208],{"href":207},"\u002Fblog\u002Fwhat-is-an-enterprise-ai-operating-system\u002F","What is an enterprise AI operating system"," is the cousin page for wiki, workstreams, and routing as a ",[211,212,213],"em",{},"product category",". This page is the runtime tests that apply to Claude Code, a LangGraph deployment, AIP, Agentforce, and Nimbus alike — then specialised by ",[193,216,218],{"href":217},"\u002Fblog\u002Finner-vs-outer-agent-harness\u002F","inner vs outer",".",[189,221,222,226,227,231,232,236],{},[193,223,225],{"href":224},"https:\u002F\u002Fwww.mckinsey.com\u002Fcapabilities\u002Fquantumblack\u002Four-insights\u002Fthe-state-of-ai","McKinsey’s 2025 State of AI"," already measured the trap: widespread use, limited scale. A fluent demo produces the first. A harness that can refuse, replay, and ratchet produces the second. ",[193,228,230],{"href":229},"https:\u002F\u002Fwww.nist.gov\u002Fitl\u002Fai-risk-management-framework\u002Fnist-ai-rmf-playbook","NIST’s AI RMF Playbook"," is the measurement language. ",[193,233,235],{"href":234},"https:\u002F\u002Fwww.iso.org\u002Fstandard\u002F42001","ISO\u002FIEC 42001"," is the management-system language. Neither is “the model seemed careful.”",[238,239,241],"h2",{"id":240},"words-youll-hear","Words you’ll hear",[243,244,245,258,274,285,295,305],"ul",{},[246,247,248,252,253,257],"li",{},[249,250,251],"strong",{},"Harness vs framework."," Library versus running loop. ",[193,254,256],{"href":255},"\u002Fblog\u002Fagent-harness-vs-agent-framework\u002F","Harness vs framework",". “We use LangChain” is not a passed test.",[246,259,260,263,264,268,269,273],{},[249,261,262],{},"Sensor."," Independent check. ",[193,265,267],{"href":266},"https:\u002F\u002Fmartinfowler.com\u002Farticles\u002Fharness-engineering.html","Böckeler","; ",[193,270,272],{"href":271},"https:\u002F\u002Fwww.thoughtworks.com\u002Fen-us\u002Finsights\u002Fblog\u002Fgenerative-ai\u002Fharness-engineering-agent-feedback-exploring-ai-coding-sensors","Thoughtworks",". Inner: tests. Outer: quote vs SoR.",[246,275,276,279,280,284],{},[249,277,278],{},"Hook \u002F interceptor."," Always runs. ",[193,281,283],{"href":282},"https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fhooks","Claude Code hooks",". Outer: fail-closed adapter.",[246,286,287,290,291,219],{},[249,288,289],{},"Quote."," Structured payload, not a paragraph. ",[193,292,294],{"href":293},"\u002Fblog\u002Fwhat-is-write-back-governance\u002F","Write-back",[246,296,297,300,301,219],{},[249,298,299],{},"Replay."," Can you reconstruct signer, policy version, tool grants. ",[193,302,304],{"href":303},"\u002Fblog\u002Fwhat-is-a-lifecycle-graph\u002F","Lifecycle graph",[246,306,307,310,311,219],{},[249,308,309],{},"Model portability."," Swap weights without rewriting tools. Not a logo on a slide. ",[193,312,314],{"href":313},"\u002Fblog\u002Fwhat-is-model-routing\u002F","Model routing",[189,316,317,318,322,323,327],{},"When you evaluate Nimbus, run these tests on ",[193,319,321],{"href":320},"\u002Fproduct\u002Fworkstreams\u002F","workstreams"," and ",[193,324,326],{"href":325},"\u002Fproduct\u002Fgovernance\u002F","governance",", not on a homepage video. When you evaluate Claude Code, run them on a repo hook and CI, not on a blog SWE-bench screenshot. Same sheet, different workspace.",[238,329,331],{"id":330},"why-evaluation-usually-fails","Why evaluation usually fails",[189,333,334],{},"People score agents like they score chat: quality of the paragraph, latency, brand of the model. That produces three false passes:",[336,337,338,348,358],"ol",{},[246,339,340,343,344,219],{},[249,341,342],{},"The copilot pass."," SSO, a usage dashboard, a good answer. No loop ownership. ",[193,345,347],{"href":346},"\u002Fblog\u002Fcollaborative-ai-and-personal-assistants\u002F","Copilot vs work OS",[246,349,350,353,354,219],{},[249,351,352],{},"The benchmark pass."," SWE-bench or Terminal-Bench for an outer job. Inner eval, outer purchase. ",[193,355,357],{"href":356},"\u002Fblog\u002Feval-loops-for-enterprise-agent-harnesses\u002F","Eval loops",[246,359,360,363,364,368],{},[249,361,362],{},"The framework pass."," A graph in a notebook with every production tool attached. ",[193,365,367],{"href":366},"https:\u002F\u002Fgenai.owasp.org\u002Fllm-top-10\u002F","OWASP"," excessive agency with extra nodes.",[189,370,371,375],{},[193,372,374],{"href":373},"https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fbuilding-effective-agents","Anthropic"," is blunt: encode the job, bound the tools, define done. Your proof of value should force those three. Written answers without a failed action are still a slide.",[189,377,378],{},"Red flags: chathe entire proof; “we integrate” with no scoped grant; governance as PDF; memory as a long window; “model-agnostic” with a flagship default and seat pricing; MCP write tools that inherit a god service account; vendor database offered as the new system of record.",[238,380,382],{"id":381},"checklist","Checklist",[189,384,385,388,389,393,394,219],{},[249,386,387],{},"1. Can it stop an action the model wants?"," Inner: ",[390,391,392],"code",{},"PreToolUse"," denies a matched command; tests fail the merge. Outer: unsigned write does not execute; reject leaves SoR unchanged. If the only stop is max tokens, you have a fuse, not a control plane. ",[193,395,397],{"href":396},"\u002Fblog\u002Fwhat-is-human-in-the-loop-ai\u002F","What is human-in-the-loop AI",[189,399,400,401,405,406,410,411,414],{},"Why this matters: ",[193,402,404],{"href":403},"https:\u002F\u002Fwww.cbc.ca\u002Fnews\u002Fcanada\u002Fbritish-columbia\u002Fair-canada-chatbot-lawsuit-1.7116416","Air Canada"," and the ",[193,407,409],{"href":408},"https:\u002F\u002Fwww.reuters.com\u002Flegal\u002Fnew-york-lawyers-sanctioned-using-fake-chatgpt-cases-legal-brief-2023-06-22\u002F","sanctioned ChatGPT brief"," are ungated generation reaching a record. Your demo must show a ",[211,412,413],{},"failed"," write.",[189,416,417,420,421,424,425,428,429,432],{},[249,418,419],{},"2. Can you replay who signed and which harness version ran?"," Signer identity, wiki or ",[390,422,423],{},"AGENTS.md"," revision, tool grants, payload hash, model class. If the answer is Slack search or “the transcript,” you do not have a ledger. ",[193,426,427],{"href":303},"What is a lifecycle graph",". Nimbus’s ",[193,430,23],{"href":431},"\u002Fproduct\u002Flifecycle-graph\u002F"," is one implementation; demand the export without a vendor engineer.",[189,434,435,438,439,442],{},[249,436,437],{},"3. Can you swap the model without rewriting tools?"," Change compact vs frontier on extract vs judgement. If tools are bound to one vendor’s function-calling dialect in application code with no adapter, portability is a hope. ",[193,440,441],{"href":202},"LangChain’s model interface"," exists for this; product harnesses must expose it as policy, not as a rewrite.",[189,444,445,448,449,453,454,219],{},[249,446,447],{},"4. Are tools grants or a belt?"," Least privilege per job. Missing Salesforce is a configuration error, not a hallucination. ",[193,450,452],{"href":451},"\u002Fintegrations\u002F","Connector architecture",". MCP servers inherit the same grant. ",[193,455,457],{"href":456},"\u002Fblog\u002Fwhat-is-model-context-protocol\u002F","What is Model Context Protocol",[189,459,460,463,464,468],{},[249,461,462],{},"5. Is verification outside the generator?"," Inner: CI the agent cannot mark skip without a hook. Outer: schema of the quote; SoR row matches. Anthropic’s ",[193,465,467],{"href":466},"https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Feffective-harnesses-for-long-running-agents","long-running harness"," refuses “premature victory” by forcing artefacts and tests. Steal that instinct.",[189,470,471,474,475,479,480,219],{},[249,472,473],{},"6. Can an operator add a sensor without a six-month SOW?"," ",[193,476,478],{"href":477},"\u002Fblog\u002Fwhat-is-harness-engineering\u002F","Harness engineering"," is a ratchet. If only vendor FDE can add a gate, you bought a programme. Fine for AIP-scale. Wrong for a standard CRM field this quarter. ",[193,481,483],{"href":482},"\u002Fevaluate\u002F","Self-service vs FDE",[189,485,486,489,490,494],{},[249,487,488],{},"7. Is the workspace the job you are buying?"," Repo vs company. ",[193,491,493],{"href":492},"\u002Fblog\u002Fhow-to-choose-between-a-coding-harness-and-an-enterprise-harness\u002F","How to choose coding vs enterprise",". A single scoring sheet with no workspace column will buy the wrong loop.",[189,496,497,500,501,505],{},[249,498,499],{},"8. Economics of the loop."," Max steps, spend cap, routing. Seat “unlimited” is often always-flagship. ",[193,502,504],{"href":503},"\u002Fblog\u002Fwhat-is-ai-token-economics\u002F","What is AI token economics",". Ask for a per-step model breakdown on a live run.",[507,508,510],"h3",{"id":509},"rfp-questions","RFP questions",[336,512,513,516,519,522,525,528,531,534],{},[246,514,515],{},"Show an action the model attempted that the harness refused. What fired?",[246,517,518],{},"After a successful write (or merge), show the signer, policy version, and payload (or diff) without Slack.",[246,520,521],{},"Change the model on extract this week. Which tools broke?",[246,523,524],{},"Attach a connector (or repo permission) as an operator, not as SE. Time?",[246,526,527],{},"Detach the grant mid-job. Does the write fail closed?",[246,529,530],{},"What is the independent sensor for “done”? Who can mark skip?",[246,532,533],{},"Two departments, different scopes, one job — or one god toolbox?",[246,535,536],{},"Price: seats, tokens, usage limits, or a services quote? What stops flagship on classify?",[189,538,539],{},"Put these in the RFP, then run them in a proof of value. Agents without a harness test are a persona list.",[507,541,543],{"id":542},"proof-of-value-short","Proof of value (short)",[189,545,546,549],{},[249,547,548],{},"Inner job:"," real repo, required hook, red test the agent must fix, no production SoR token.",[189,551,552,555,556,559],{},[249,553,554],{},"Outer job:"," real cross-department write, quoted payload, reject path, export. Nimbus should pass the same live sequence as anyone else: OAuth attach, blocked unsigned write, graph export. ",[193,557,11],{"href":558},"\u002Foverview\u002F"," is not the proof.",[189,561,562],{},"Skip any refuse\u002Freplay\u002Fswap and you evaluated a chat product, a benchmark, or a framework notebook.",[238,564,566],{"id":565},"score-inner-and-outer-without-mixing-oracles","Score inner and outer without mixing oracles",[189,568,569],{},"Run two short scripts. Do not average them into one “AI score.”",[189,571,572,575,576,578],{},[249,573,574],{},"Inner script (repo)."," Fresh checkout of a service you own. Required hook: deny a dangerous bash pattern. Agent must add a failing test then make it pass. CI is the merge sensor. No production CRM token in the environment. Record: did the hook fire, did CI stay independent, can you show the ",[390,577,423],{}," revision. SWE-bench plots from the vendor are background, not this script.",[189,580,581,584],{},[249,582,583],{},"Outer script (SoR)."," Sandbox Salesforce or equivalent. Operator (not SE) attaches OAuth. Model proposes a write. Unsigned path must fail. Reject path must leave records unchanged. Approve path: read-back matches hash. Export signer and wiki revision. Detach the connector and retry the write — must fail closed. Proof of value is this script with two departments on the canvas.",[189,586,587,588,322,590,592,593,219],{},"If a vendor refuses to run the outer script because “we are a coding tool,” believe them and buy them for inner only. If a vendor refuses the inner script because “we are an OS,” believe them and do not replace Cursor. If a vendor claims both and fails one script, you have a category error in their marketing. Nimbus should pass the outer script on ",[193,589,321],{"href":320},[193,591,326],{"href":325},". Claude Code should pass the inner script. ",[193,594,595],{"href":492},"How to choose",[189,597,598,601,602,605],{},[249,599,600],{},"Thoughtworks’ layer check."," After the scripts, ask where layer 4 lives: who owns the policy when the agent did what it was allowed to do and harm still happened. If the answer is a steering committee with no interceptor, you evaluated theatre. ",[193,603,604],{"href":234},"ISO 42001"," will not save a missing refuse.",[189,607,608,611,612,219],{},[249,609,610],{},"Economics check."," Pull one live run’s step list: model class per step, tokens or cost, which sensor fired. Always-flagship with no cap is a failed harness eval even if the paragraph was good. ",[193,613,614],{"href":503},"Cost control",[189,616,617,620,621,219],{},[249,618,619],{},"MCP check."," One write-capable server. Which workspaces may use it. If the answer is “any host that can see the URL,” fail. ",[193,622,457],{"href":456},[189,624,625,626,630],{},"Weight the eight checklist items; do not add a ninth called “brand.” ",[193,627,629],{"href":628},"https:\u002F\u002Fhai.stanford.edu\u002Fai-index\u002F2025-ai-index-report","Stanford AI Index"," is useful context for how fast coding tools moved. It is not a substitute for the outer script.",[189,632,633,634,638,639,642],{},"Score vendorsystems, not as essays. A beautiful ",[193,635,637],{"href":636},"https:\u002F\u002Fwww.langchain.com\u002Fblog\u002Fthe-anatomy-of-an-agent-harness","anatomy post"," does not pass the refuse test. A messy UI that blocks the unsigned PATCH does. Watch for “evaluation theatre”: the SE runs the happy path, the fail path is “we’ll configure that in phase two,” the ledger is a screenshot of LangSmith. Phase two is where ",[193,640,641],{"href":224},"McKinsey"," pilots go to die.",[189,644,645,646,649],{},"Bring your own oracle. For inner: a test the agent did not write. For outer: a sandbox row you control. If the vendor must supply the only success criterion, you are scoring their demo fixtures. Terminal-Bench’s strength is that the ",[211,647,648],{},"environment"," is the grader. Copy that.",[189,651,652],{},"People on the bake-off: an operator who will live in the product, someone who owns the SoR, someone who can say no for Legal, an engineer who will keep the inner harness. If only the vendor and an innovation lead attend, you will buy a narrative. Nimbus, AIP, Cursor, and a LangGraph SOW should all survive that room or be narrowed to the job they actually do.",[189,654,655,656,659],{},"Write the pass\u002Ffail before the demo so the SE cannot redefine success live. “Blocked unsigned write” is a boolean. “Felt enterprise-ready” is not. Record the session. If they cannot fail on camera, assume they cannot fail in production. ",[193,657,658],{"href":229},"NIST Playbook"," language helps here: you are Measuring a control, not a vibe.",[238,661,663],{"id":662},"how-this-shows-up-in-nimbus","How this shows up in Nimbus",[189,665,666,667,671,672,676,677,681,682,684,685,689],{},"Nimbus is an ",[193,668,670],{"href":669},"\u002Fblog\u002Fwhat-is-an-enterprise-agent-harness\u002F","enterprise \u002F outer harness",": ",[193,673,675],{"href":674},"\u002Fproduct\u002Fwiki\u002F","wiki"," as guides, connectors as grants, ",[193,678,680],{"href":679},"\u002Fproduct\u002Fagent-teams\u002F","teams"," as the hiring object, ",[193,683,326],{"href":325}," as the interceptor, graph as replay, ",[193,686,688],{"href":687},"\u002Fmodels\u002F","models"," as routing. Score those surfaces against the eight tests. Do not accept “we are a harness” as a substitute for a failed write. AIP and Agentforce deserve the same eight.",[238,691,693],{"id":692},"questions-people-actually-ask","Questions people actually ask",[507,695,697],{"id":696},"can-we-score-claude-code-and-nimbus-on-one-spreadsheet","Can we score Claude Code and Nimbus on one spreadsheet?",[189,699,700],{},"Yes, with a workspace column. Shared rows: refuse, replay, swap, sensors, operator change, economics. Inner-only rows: tests, sandbox, PR. Outer-only rows: SoR quote, roster signer, workstream isolation.",[507,702,704],{"id":703},"the-vendor-sent-a-swe-bench-plot","The vendor sent a SWE-bench plot.",[189,706,707,708,219],{},"File it under inner quality. If you are buying CRM writes, it is not sufficient. ",[193,709,357],{"href":356},[507,711,713],{"id":712},"we-already-completed-a-copilot-rfp","We already completed a copilot RFP.",[189,715,716],{},"Keep it for personal tools. This sheet is for loops that act. Different job.",[507,718,720],{"id":719},"is-iso-42001-certification-the-eval","Is ISO 42001 certification the eval?",[189,722,723],{},"It is a management-system signal. Still watch a write fail. Certification without an interceptor is paperwork.",[507,725,727],{"id":726},"what-should-i-read-next","What should I read next?",[189,729,730,734,735,738,739,742],{},[193,731,733],{"href":732},"\u002Fblog\u002Fagent-harness-architecture\u002F","Agent harness architecture"," to know the parts. ",[193,736,737],{"href":293},"What is write-back governance"," for the outer stop in detail. ",[193,740,741],{"href":477},"What is harness engineering"," for the ratchet after you buy.",[238,744,746],{"id":745},"related-reading","Related reading",[189,748,749,322,753,219],{},[193,750,752],{"href":751},"\u002Fblog\u002Fwhat-is-multi-agent-ai\u002F","What is multi-agent AI",[193,754,756],{"href":755},"\u002Fblog\u002Fwhat-is-ai-governance\u002F","what is AI governance",[238,758,760],{"id":759},"sources","Sources",[243,762,763,768,773,778,783,788,793,798,803,808,812,817,822,827,832],{},[246,764,765],{},[193,766,767],{"href":202},"LangChain, Agents",[246,769,770],{},[193,771,772],{"href":636},"LangChain, The anatomy of an agent harness",[246,774,775],{},[193,776,777],{"href":266},"Böckeler, Harness engineering for coding agent users",[246,779,780],{},[193,781,782],{"href":271},"Thoughtworks, Harness engineering and agent feedback",[246,784,785],{},[193,786,787],{"href":373},"Anthropic, Building effective agents",[246,789,790],{},[193,791,792],{"href":466},"Anthropic, Effective harnesses for long-running agents",[246,794,795],{},[193,796,797],{"href":282},"Claude Code, Hooks",[246,799,800],{},[193,801,802],{"href":224},"McKinsey, The state of AI in 2025",[246,804,805],{},[193,806,807],{"href":229},"NIST AI RMF Playbook",[246,809,810],{},[193,811,235],{"href":234},[246,813,814],{},[193,815,816],{"href":366},"OWASP Top 10 for LLM applications",[246,818,819],{},[193,820,821],{"href":403},"CBC, Air Canada chatbot lawsuit",[246,823,824],{},[193,825,826],{"href":408},"Reuters, ChatGPT legal brief sanctions",[246,828,829],{},[193,830,831],{"href":628},"Stanford HAI, 2025 AI Index",[246,833,834],{},[193,835,837],{"href":836},"https:\u002F\u002Fmodelcontextprotocol.io\u002Fspecification\u002F2025-11-25\u002Findex","Model Context Protocol specification",{"title":169,"searchDepth":170,"depth":170,"links":839},[840,841,842,847,848,849,856,857],{"id":240,"depth":170,"text":241},{"id":330,"depth":170,"text":331},{"id":381,"depth":170,"text":382,"children":843},[844,846],{"id":509,"depth":845,"text":510},3,{"id":542,"depth":845,"text":543},{"id":565,"depth":170,"text":566},{"id":662,"depth":170,"text":663},{"id":692,"depth":170,"text":693,"children":850},[851,852,853,854,855],{"id":696,"depth":845,"text":697},{"id":703,"depth":845,"text":704},{"id":712,"depth":845,"text":713},{"id":719,"depth":845,"text":720},{"id":726,"depth":845,"text":727},{"id":745,"depth":170,"text":746},{"id":759,"depth":170,"text":760},"2026-08-24","Evaluating an agent harness means checking whether it can stop a write, replay who signed, and swap the model without rewriting the tools.",{"eyebrow":861,"title":862},"Short answers","Stops, replay, and a real sensor",[864,867,870],{"question":865,"answer":866},"What should an evaluation of a harness check?","Whether it can stop a write, replay who signed, swap the model without rewriting the tools, and fail a real sensor.",{"question":868,"answer":869},"Is a good demo answer enough?","No. A fluent answer does not prove the harness can refuse a bad change.",{"question":871,"answer":872},"What is a failed evaluation?","The write still lands when the check fails, or nobody can reconstruct who approved last Tuesday's change.","\u002Fblog\u002Fhow-to-evaluate-an-agent-harness",{"title":179,"description":859},"evaluation","blog\u002Fhow-to-evaluate-an-agent-harness",[875,878,326,879],"agent-harness","rfp","Gv9m9kXME8wRmG6-l6R6IYqmcl4wOhGLVhOAjNzk4xU",{"hero":882,"id":884,"title":885,"archived":163,"authors":164,"badge":164,"body":886,"date":164,"definedTerm":164,"department":164,"description":890,"extension":172,"eyebrow":891,"faqHeader":164,"faqs":164,"footerBand":892,"headline":164,"image":164,"industry":164,"jobType":164,"listed":131,"location":164,"navigation":131,"openRoles":164,"pageLayout":164,"path":60,"relatedHeading":898,"seo":899,"series":164,"sitemap":131,"status":164,"stem":900,"subhead":164,"tags":164,"video":164,"whyJoin":164,"workplaceType":164,"__hash__":901},{"filename":883},"u2221455217_Flat_design_of_a_futuristic_minimalist_landscape__5d589295-cdea-4ea9-a262-be766881accf_1.png","content\u002Fblog\u002Findex.md","Exploring the future of intelligence.",{"type":166,"value":887,"toc":888},[],{"title":169,"searchDepth":170,"depth":170,"links":889},[],"Deep dives into pre-cognitive intelligence, sentient enterprises, and the evolving landscape of AI-driven business transformation.","Latest Research",{"headline":893,"description":894,"primaryLabel":895,"primaryTo":896,"secondaryLabel":897,"secondaryTo":12},"Stay at the frontier.","Subscribe for product updates and new insights.","Subscribe","\u002Fnewsletter","Explore the platform","More research",{"title":885,"description":890},"blog\u002Findex","BFSWGYO9bcTlaulivKYWyg08_DJHsdGg3OC6g_CG1Hw",[903,1537],{"id":904,"title":905,"archived":163,"authors":906,"badge":908,"body":910,"date":858,"definedTerm":164,"department":164,"description":1516,"extension":172,"eyebrow":164,"faqHeader":1517,"faqs":1519,"footerBand":164,"headline":164,"image":164,"industry":164,"jobType":164,"listed":131,"location":164,"navigation":131,"openRoles":164,"pageLayout":164,"path":1529,"relatedHeading":164,"seo":1530,"series":1531,"sitemap":131,"status":164,"stem":1532,"subhead":164,"tags":1533,"video":164,"whyJoin":164,"workplaceType":164,"__hash__":1536},"content\u002Fblog\u002Finner-vs-outer-agent-harness.md","Inner vs Outer Agent Harness",[907],{"name":182,"to":183},{"label":909},"Explainer",{"type":166,"value":911,"toc":1499},[912,927,944,965,979,981,1055,1076,1080,1083,1086,1100,1114,1117,1139,1149,1153,1169,1179,1189,1197,1207,1217,1235,1244,1259,1263,1266,1276,1284,1293,1297,1307,1316,1325,1330,1335,1345,1348,1350,1354,1357,1361,1368,1372,1382,1386,1391,1395,1401,1403,1417,1419,1425,1427],[189,913,914,915,918,919,922,923,926],{},"An ",[249,916,917],{},"inner agent harness"," is the runtime around a model for a developer and a codebase. An ",[249,920,921],{},"outer agent harness"," is the runtime around a model for operators and live business systems. Same equation — ",[193,924,925],{"href":202},"Agent = Model + Harness"," — different workspace, different sensors, different stop.",[189,928,929,931,932,935,936,940,941,219],{},[193,930,267],{"href":266}," already uses “outer harness” for the controls ",[211,933,934],{},"users"," add around a coding agent (guides, sensors) as distinct from the vendor’s built-in loop. ",[193,937,939],{"href":938},"https:\u002F\u002Faddyosmani.com\u002Fblog\u002Fown-the-outer-loop\u002F","Addy Osmani"," tells engineers to own the outer loop of investigate → implement → verify so accountability does not dissolve into the model. This article borrows those words and draws the cut enterprises actually buy: ",[249,942,943],{},"repo versus company",[189,945,946,947,950,951,954,955,950,957,960,961,964],{},"Claude Code, Cursor, and Codex are excellent inner harnesses. They sandboxes, ",[390,948,949],{},"apply_patch",", ",[390,952,953],{},"CLAUDE.md"," \u002F ",[390,956,423],{},[193,958,959],{"href":282},"hooks",", and tests. Palantir AIP, Salesforce Agentforce, and OS-class products such as Nimbus are outer harnesses: ",[193,962,321],{"href":963},"\u002Fblog\u002Fwhat-is-an-ai-workstream\u002F",", connectors, named signers, a decision record. Confusing them is how Legal is asked to “just use Cursor on the Salesforce repo” and how engineering is asked to “approve CRM writes in a coding agent.”",[189,966,967,970,971,974,975,978],{},[193,968,969],{"href":492},"How to choose between a coding harness and an enterprise harness"," is the buying version of this page. ",[193,972,973],{"href":346},"Collaborative AI and personal assistants"," is the adjacent cut (personal assistant versus departmental work). Inner\u002Fouter is about ",[211,976,977],{},"which loop you are hiring",", not whether the UI looks like chat.",[238,980,241],{"id":240},[243,982,983,989,995,1005,1018,1027,1041],{},[246,984,985,988],{},[249,986,987],{},"Inner loop (classic SE)."," Edit, build, test on a developer’s machine. Fast. Local. The coding-agent inner harness lives here: shell, files, compiler.",[246,990,991,994],{},[249,992,993],{},"Outer loop (classic SE)."," PR, CI, review, release. Osmani’s “own the outer loop” is this accountability layer for agentic coding. Still software.",[246,996,997,1000,1001,1004],{},[249,998,999],{},"Inner harness (this article)."," Vendor + user controls for a ",[249,1002,1003],{},"repository workspace",": Claude Code, Cursor, Codex. Eval: tests, Terminal-Bench, SWE-bench.",[246,1006,1007,1010,1011,1014,1015,219],{},[249,1008,1009],{},"Outer harness (this article)."," Controls for a ",[249,1012,1013],{},"company workspace",": jobs, systems of record, people who may sign. Eval: quoted write, identity, ledger. An ",[193,1016,1017],{"href":669},"enterprise agent harness",[246,1019,1020,1023,1024,219],{},[249,1021,1022],{},"Guides vs sensors."," Feed-forward markdown versus feedback from tools. Inner: lint and pytest. Outer: schema of a Salesforce payload and a Hard gate. See ",[193,1025,1026],{"href":477},"what is harness engineering",[246,1028,1029,1032,1033,1036,1037,219],{},[249,1030,1031],{},"CLAUDE.md \u002F AGENTS.md."," Inner guides. ",[193,1034,374],{"href":1035},"https:\u002F\u002Fclaude.com\u002Fblog\u002Fsteering-claude-code-skills-hooks-rules-subagents-and-more"," is explicit: files are context; hooks are deterministic. A company wiki is the outer analogue of those files — asserted policy, not a repo README. See ",[193,1038,1040],{"href":1039},"\u002Fblog\u002Fwhat-is-a-company-wiki-for-ai-agents\u002F","What is a company wiki for AI agents",[246,1042,1043,1046,1047,1050,1051,1054],{},[249,1044,1045],{},"Write gate."," Inner: hook denies ",[390,1048,1049],{},"rm"," or force-push. Outer: ",[193,1052,1053],{"href":293},"write-back governance"," — adapter cannot mutate CRM until a named role signs the quote.",[189,1056,1057,1058,1060,1061,950,1063,1066,1067,1069,1070,1060,1072,1075],{},"Nimbus is built as an outer harness: ",[193,1059,675],{"href":674}," instead of only ",[390,1062,423],{},[193,1064,1065],{"href":451},"connectors"," instead of only a local shell, ",[193,1068,326],{"href":325}," instead of only a pre-commit hook, ",[193,1071,23],{"href":431},[390,1073,1074],{},"git log",". Engineering should still run Claude Code. Those products should not share a write path to NetSuite.",[238,1077,1079],{"id":1078},"why-you-should-care","Why you should care",[189,1081,1082],{},"Demos collapse the cut. Both products answer a question. Both call tools. Both show a transcript. The evaluation is the workspace.",[189,1084,1085],{},"It affects you if:",[243,1087,1088,1091,1094,1097],{},[246,1089,1090],{},"Security asks whether the coding agent’s MCP server can reach production Salesforce",[246,1092,1093],{},"RevOps wants “an agent” and is shown a SWE-bench slide",[246,1095,1096],{},"Engineering wants Cursor and is told to wait for the enterprise OS",[246,1098,1099],{},"You already have both, and they silently write to the same object",[189,1101,1102,1104,1105,1108,1109,1113],{},[193,1103,225],{"href":224}," describes agentic systems as an organisational design problem. Inner harnesses scale developer throughput. They do not, by themselves, scale governed operations. ",[193,1106,1107],{"href":628},"Stanford HAI’s 2025 AI Index"," maps how fast coding-agent tooling moved. Speed in the repo is not a substitute for ",[193,1110,1112],{"href":1111},"https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2024\u002F1689\u002Foj","EU AI Act"," oversight on systems that affect customers and money.",[189,1115,1116],{},"Two failure modes:",[336,1118,1119,1129],{},[246,1120,1121,1124,1125,1128],{},[249,1122,1123],{},"Outer job, inner harness."," A pricing change drafted in Cursor with an MCP Salesforce tool. Tests pass on a fixture. Production Amount changes. ",[390,1126,1127],{},"git blame"," does not name the signer. You used a repo loop on a company record.",[246,1130,1131,1134,1135,1138],{},[249,1132,1133],{},"Inner job, outer harness."," “Rewrite this function” opened as a cross-department ",[193,1136,1137],{"href":320},"workstream"," with a Critical gate. Engineers will route around it. You used a company loop on a compile.",[189,1140,1141,1145,1146,1148],{},[193,1142,1144],{"href":1143},"https:\u002F\u002Fwww.nist.gov\u002Fitl\u002Fai-risk-management-framework","NIST AI RMF"," Map step: know the context of use. Inner and outer are different contexts. ",[193,1147,235],{"href":234}," wants controls matched to that context. One harness policy for “all AI” is how both jobs get the wrong stop.",[238,1150,1152],{"id":1151},"what-each-harness-actually-owns","What each harness actually owns",[189,1154,1155,1158,1159,1161,1162,1165,1166,1168],{},[249,1156,1157],{},"Workspace."," Inner: a checkout, often sandboxed. Anthropic’s ",[193,1160,467],{"href":466}," keeps progress in git and files because the workspace ",[211,1163,1164],{},"is"," the filesystem. Outer: a job folder with people, budget, and attached systems — a ",[193,1167,1137],{"href":963},". Files may appear as artefacts. They are not the system of record.",[189,1170,1171,1174,1175,1178],{},[249,1172,1173],{},"Identity."," Inner: the developer’s machine credentials, a repo token, maybe a sandbox role. Outer: org roster, workstream membership, named approver. The model is not the principal. ",[193,1176,1177],{"href":451},"Connector and permissions architecture"," is the outer identity plane.",[189,1180,1181,1184,1185,1188],{},[249,1182,1183],{},"Tools."," Inner: shell, editor, tests, browser, maybe MCP to docs. Outer: CRM, ERP, warehouse, ticket systems, mail — default read, write as a separate plane. ",[193,1186,1187],{"href":456},"MCP"," can sit under both. The grant must not.",[189,1190,1191,388,1194,1196],{},[249,1192,1193],{},"Guides.",[390,1195,423],{},", skills, directory-local rules. Outer: company wiki, playbooks versioned with the run. Mixing them is useful (engineering conventions in the repo; discount policy in the wiki). Collapsing them is how a style guide becomes “legal approval.”",[189,1198,1199,1202,1203,1206],{},[249,1200,1201],{},"Sensors."," Inner: typechecker, unit tests, CI, architecture tests. Böckeler and ",[193,1204,1205],{"href":271},"Thoughtworks on sensors",". Outer: payload schema, blast-radius cardinality, maker-checker, exportable ledger. A passing pytest does not mean Opportunity.Stage was authorised.",[189,1208,1209,1212,1213,1216],{},[249,1210,1211],{},"Stop."," Inner: tests red, hook exit 2, max steps, human in the IDE. Outer: wait-for-named-signer, missing connector, budget, reject. ",[193,1214,1215],{"href":396},"Human-in-the-loop"," in a coding agent is “the developer kept going.” HITL in an outer harness is a first-class step with identity.",[189,1218,1219,388,1222,950,1226,1230,1231,1234],{},[249,1220,1221],{},"Eval.",[193,1223,1225],{"href":1224},"https:\u002F\u002Fwww.swebench.com\u002F","SWE-bench",[193,1227,1229],{"href":1228},"https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.11868","Terminal-Bench",", your suite. Outer: replay the signer; compare quote to SoR; see ",[193,1232,1233],{"href":356},"eval loops",". Leaderboard scores are not a SOX control.",[189,1236,1237,1240,1241,1243],{},[249,1238,1239],{},"Memory."," Inner: files, commits, session transcripts, memory files the next coding session loads. Outer: wiki + ",[193,1242,23],{"href":303}," so next quarter’s operator can ask why a field changed. Chat logs of a coding session are not institutional memory for RevOps.",[189,1245,1246,1247,1250,1251,1254,1255,1258],{},"Nimbus’s ",[193,1248,1249],{"href":679},"agent teams"," sit on the outer side: mandates, required connectors, approval triggers. You can still ",[211,1252,1253],{},"use"," an inner harness as a bounded tool behind a connector (for example a coding agent that only opens a draft PR). Do not let that inner harness become the orchestrator of record for a CRM write. ",[193,1256,1257],{"href":751},"Multi-agent architecture"," says the same thing with specialists: hands are not roles.",[238,1260,1262],{"id":1261},"how-they-should-sit-together","How they should sit together",[189,1264,1265],{},"Most companies need both. That is not a hedge. It is how software and operations already split.",[189,1267,1268,1271,1272,1275],{},[249,1269,1270],{},"Pattern that works."," Engineers use Cursor or Claude Code on application repos. CI remains the merge sensor. Separately, RevOps and Finance run outer-harness jobs on Salesforce and NetSuite. If a coding agent must touch a live business system, it proposes an artefact; the outer harness quotes and gates the write. Two writers to the same object without a single quote is the failure ",[193,1273,1274],{"href":751},"multi-agent architecture"," already names.",[189,1277,1278,1281,1282,219],{},[249,1279,1280],{},"Pattern that fails."," One MCP mesh with production tokens, used from the IDE and from the chatbot and from the OS. Confused deputy. ",[193,1283,457],{"href":456},[189,1285,1286,1289,1290,219],{},[249,1287,1288],{},"Thoughtworks’ four layers"," — model, builder harness, user harness, organisational harness — map cleanly: Claude Code is builder + user on the inner side; the organisational layer is the outer operating model. Nimbus is one productisation of that outer layer, not the only one. AIP is a programme-shaped outer harness. Agentforce is CRM-anchored. Score scope and time-to-value separately. See ",[193,1291,1292],{"href":482},"self-service vs forward-deployed",[238,1294,1296],{"id":1295},"a-week-that-uses-both","A week that uses both",[189,1298,1299,1300,1302,1303,1306],{},"Monday an engineer uses Cursor to fix a pricing calculator in the billing service. ",[390,1301,423],{}," says no raw SQL in the request path. A hook blocks ",[390,1304,1305],{},"git push --force",". CI runs the unit suite. The PR is the artefact. CODEOWNERS signs the merge. That is a complete inner story. SWE-bench is relevant only as a vendor quality signal for the coding tool, not as a control.",[189,1308,1309,1310,1312,1313,1315],{},"Tuesday RevOps needs the list price on twenty renewals updated after Legal changed the cap in the playbook. The artefact is Salesforce. The signer is a named RevOps lead. The sensor is: quoted fields, hash, read-back. If Tuesday’s job is opened as a Cursor session with an MCP Salesforce server using a shared integration user, you have imported Monday’s workspace into Tuesday’s system of record. ",[390,1311,1074],{}," will not name the RevOps lead. ",[193,1314,294],{"href":293}," did not fire because the inner harness does not have that interceptor.",[189,1317,1318,1319,1321,1322,1324],{},"Wednesday someone proposes “one agent for everything.” The honest architecture is: Monday’s harness stays. Tuesday’s job runs on an outer harness — in Nimbus, a ",[193,1320,1137],{"href":320}," with the CRM connector, the wiki revision that contains the new cap, a Hard gate. If the calculator ",[211,1323,390],{}," must change as well, the outer job can spawn a bounded inner step that opens a draft PR. Two artefacts, two sensors, one company rule: unsigned SoR writes are impossible from either loop.",[189,1326,1327,1328,219],{},"Thursday Security reviews MCP. The question is not “is MCP approved.” It is “which workspace may this server mutate.” Inner: sandbox and repo. Outer: workstream grant. Same protocol, different identity box. ",[193,1329,457],{"href":456},[189,1331,1332,1333,219],{},"Friday you look at evals. Engineering posts a Terminal-Bench plot for the coding vendor. Finance asks who signed Amount. Those are not competing dashboards. They are different oracles. ",[193,1334,357],{"href":356},[189,1336,1337,1340,1341,1344],{},[193,1338,272],{"href":1339},"https:\u002F\u002Fwww.thoughtworks.com\u002Finsights\u002Farticles\u002Foperating-system-enterprise-ai"," would call Monday layers 2–3 on a builder harness, Tuesday a delegation question on layer 4, and “one agent” a way to skip layer 4. ",[193,1342,1343],{"href":938},"Osmani"," would say engineering still owns verify-and-merge on Monday. Neither author is selling Nimbus. Both are describing why the cut exists.",[189,1346,1347],{},"If you only fund inner harnesses, Tuesday happens in paste and Slack. If you only fund outer harnesses, Monday happens in unsanctioned Cursor anyway. Fund both. Bind writes.",[238,1349,693],{"id":692},[507,1351,1353],{"id":1352},"is-cursor-an-enterprise-harness-if-we-sso-it","Is Cursor an enterprise harness if we SSO it?",[189,1355,1356],{},"SSO is admin control. It does not quote a NetSuite journal or bind a Finance signer. Cursor can be an inner harness in an enterprise. That is not the same as an outer harness.",[507,1358,1360],{"id":1359},"can-claude-code-hooks-replace-write-back-governance","Can Claude Code hooks replace write-back governance?",[189,1362,1363,1364,1367],{},"They can replace ",[211,1365,1366],{},"some"," inner invariants (dangerous bash). They do not give you a payload in the language of Salesforce, a roster-aware approver, or an exportable operations ledger. Different workspace.",[507,1369,1371],{"id":1370},"should-we-ban-coding-agents-until-the-os-is-live","Should we ban coding agents until the OS is live?",[189,1373,1374,1375,1378,1379,219],{},"Usually no. Ban unsigned writes to systems of record from ",[211,1376,1377],{},"any"," agent, inner or outer. Let inner harnesses keep compiling. ",[193,1380,1381],{"href":293},"How to solve unapproved CRM writes from AI",[507,1383,1385],{"id":1384},"where-does-a-copilot-fit","Where does a copilot fit?",[189,1387,1388,1389,219],{},"A copilot is often not a full inner harness — no repo loop, no tests. Personal throughput. Keep it for mail. Do not give it the CRM write token. ",[193,1390,347],{"href":346},[507,1392,1394],{"id":1393},"is-nimbus-trying-to-replace-claude-code","Is Nimbus trying to replace Claude Code?",[189,1396,1397,1398,1400],{},"No. Different workspace. Nimbus is the company loop; Claude Code is the repo loop. ",[193,1399,11],{"href":558}," is the product map. This page is the architectural cut.",[507,1402,727],{"id":726},[189,1404,1405,1408,1409,1412,1413,1416],{},[193,1406,1407],{"href":669},"What is an enterprise agent harness",". ",[193,1410,1411],{"href":255},"Agent harness vs agent framework"," if you are assembling rather than hiring. ",[193,1414,1415],{"href":195},"What is an agent harness"," for the base noun.",[238,1418,746],{"id":745},[189,1420,1421,322,1423,219],{},[193,1422,741],{"href":477},[193,1424,733],{"href":732},[238,1426,760],{"id":759},[243,1428,1429,1433,1437,1442,1448,1453,1457,1462,1466,1470,1475,1479,1483,1487,1491,1495],{},[246,1430,1431],{},[193,1432,767],{"href":202},[246,1434,1435],{},[193,1436,777],{"href":266},[246,1438,1439],{},[193,1440,1441],{"href":938},"Addy Osmani, Own the outer loop",[246,1443,1444],{},[193,1445,1447],{"href":1446},"https:\u002F\u002Faddyosmani.com\u002Fblog\u002Fagent-harness-engineering\u002F","Addy Osmani, Agent harness engineering",[246,1449,1450],{},[193,1451,1452],{"href":1339},"Thoughtworks, The operating system for enterprise AI",[246,1454,1455],{},[193,1456,792],{"href":466},[246,1458,1459],{},[193,1460,1461],{"href":1035},"Anthropic, Steering Claude Code",[246,1463,1464],{},[193,1465,797],{"href":282},[246,1467,1468],{},[193,1469,1225],{"href":1224},[246,1471,1472],{},[193,1473,1474],{"href":1228},"Terminal-Bench (arXiv:2601.11868)",[246,1476,1477],{},[193,1478,802],{"href":224},[246,1480,1481],{},[193,1482,831],{"href":628},[246,1484,1485],{},[193,1486,1144],{"href":1143},[246,1488,1489],{},[193,1490,235],{"href":234},[246,1492,1493],{},[193,1494,1112],{"href":1111},[246,1496,1497],{},[193,1498,837],{"href":836},{"title":169,"searchDepth":170,"depth":170,"links":1500},[1501,1502,1503,1504,1505,1506,1514,1515],{"id":240,"depth":170,"text":241},{"id":1078,"depth":170,"text":1079},{"id":1151,"depth":170,"text":1152},{"id":1261,"depth":170,"text":1262},{"id":1295,"depth":170,"text":1296},{"id":692,"depth":170,"text":693,"children":1507},[1508,1509,1510,1511,1512,1513],{"id":1352,"depth":845,"text":1353},{"id":1359,"depth":845,"text":1360},{"id":1370,"depth":845,"text":1371},{"id":1384,"depth":845,"text":1385},{"id":1393,"depth":845,"text":1394},{"id":726,"depth":845,"text":727},{"id":745,"depth":170,"text":746},{"id":759,"depth":170,"text":760},"An inner agent harness runs a developer and a repository — CLAUDE.md, hooks, tests.",{"eyebrow":861,"title":1518},"The repo and the company",[1520,1523,1526],{"question":1521,"answer":1522},"What is an inner agent harness?","The runtime for a developer and a repository: CLAUDE.md, hooks, and tests.",{"question":1524,"answer":1525},"What is an outer harness?","The runtime for the company: wiki, connectors, write gates, and a ledger of what changed.",{"question":1527,"answer":1528},"Can one replace the other?","No. Most enterprises need both. A perfect repo harness does not stop an unapproved CRM write, and a company gate does not run the unit tests.","\u002Fblog\u002Finner-vs-outer-agent-harness",{"title":905,"description":1516},"explainer","blog\u002Finner-vs-outer-agent-harness",[1531,878,1534,1535],"coding-agents","enterprise-ai","03H2FWvQKQ3zCeocBdx6w-GVtDt-Nz1nBE0Sz22oZ2I",{"id":1538,"title":1539,"archived":163,"authors":1540,"badge":1542,"body":1543,"date":858,"definedTerm":164,"department":164,"description":2072,"extension":172,"eyebrow":164,"faqHeader":2073,"faqs":2075,"footerBand":164,"headline":164,"image":164,"industry":164,"jobType":164,"listed":131,"location":164,"navigation":131,"openRoles":164,"pageLayout":164,"path":2085,"relatedHeading":164,"seo":2086,"series":1531,"sitemap":131,"status":164,"stem":2087,"subhead":164,"tags":2088,"video":164,"whyJoin":164,"workplaceType":164,"__hash__":2091},"content\u002Fblog\u002Fagent-harness-vs-agent-framework.md","Agent Harness vs Agent Framework",[1541],{"name":182,"to":183},{"label":909},{"type":166,"value":1544,"toc":2055},[1545,1554,1581,1597,1600,1602,1651,1664,1666,1672,1674,1688,1702,1707,1711,1717,1731,1746,1759,1764,1774,1784,1797,1804,1808,1836,1842,1846,1856,1866,1872,1877,1886,1902,1909,1914,1917,1923,1925,1929,1936,1940,1946,1950,1953,1957,1960,1964,1971,1975,1983,1985,1992,1994],[189,1546,914,1547,1550,1551,1553],{},[249,1548,1549],{},"agent framework"," is a library for composing models, tools, and control flow. An ",[249,1552,196],{}," is the running environment around a model: the loop, the tools as they are actually granted, the stops, the sensors, and the identity that production will use.",[189,1555,1556,1559,1560,474,1563,1566,1567,1571,1572,1576,1577,1580],{},[193,1557,1558],{"href":202},"LangChain’s own docs"," are careful with the words. ",[249,1561,1562],{},"Agent = Model + Harness.",[390,1564,1565],{},"create_agent"," is “a highly configurable harness.” ",[193,1568,1570],{"href":1569},"https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Fdeepagents","Deep Agents"," is “the batteries-included agent harness.” ",[193,1573,1575],{"href":1574},"https:\u002F\u002Fdocs.langchain.com\u002Foss\u002Fpython\u002Flanggraph\u002Foverview","LangGraph"," is the low-level orchestration framework when the built-in loop is the wrong shape. That taxonomy is the whole article: a framework can ",[211,1578,1579],{},"implement"," a harness. Shipping the pip package does not mean you have one operators can hire.",[189,1582,1583,1587,1588,1590,1591,1593,1594,1596],{},[193,1584,1586],{"href":1585},"https:\u002F\u002Fwww.langchain.com\u002Fblog\u002Fhow-to-build-a-custom-agent-harness","LangChain’s custom-harness post"," says the same from the other side. Pre-assembled harnesses (Deep Agents, Claude Agent SDK) get you to a working agent fast. ",[390,1589,1565],{}," is minimal on purpose: core loop plus middleware. You still choose tools, guardrails, and business logic. CrewAI, Semantic Kernel, AutoGen, and Pydantic AI live in this neighbourhood. They are how engineers assemble loops. They are not a substitute for ",[193,1592,1053],{"href":293},", a ",[193,1595,1137],{"href":963},", or a ledger.",[189,1598,1599],{},"Claude Code and Cursor are harnesses you run, not frameworks you import. Nimbus, Palantir AIP, and Agentforce are (different) harnesses you run for company jobs. Confusing “we use LangGraph” with “we have an enterprise harness” is the 2026 version of “we use Kubernetes” meaning “we have a product.”",[238,1601,241],{"id":240},[243,1603,1604,1610,1622,1630,1636,1642],{},[246,1605,1606,1609],{},[249,1607,1608],{},"Framework."," SDKs and graphs: LangChain, LangGraph, CrewAI, AutoGen, Semantic Kernel, Pydantic AI. You write code. You own production identity unless you add it.",[246,1611,1612,1615,1616,1618,1619,1621],{},[249,1613,1614],{},"Harness."," Runtime around the model. ",[193,1617,1415],{"href":195},". May be a product (Claude Code) or a configured framework (your ",[390,1620,1565],{}," plus hooks plus IdP).",[246,1623,1624,1627,1628,219],{},[249,1625,1626],{},"Middleware \u002F hooks."," Framework primitive that becomes harness behaviour when it always runs. LangChain middleware; ",[193,1629,283],{"href":282},[246,1631,1632,1635],{},[249,1633,1634],{},"Batteries-included harness."," Deep Agents, Claude Agent SDK, Codex SDK. Opinionated loop, filesystem, subagents, compaction. Still not your CRM grant model.",[246,1637,1638,1641],{},[249,1639,1640],{},"Orchestration framework."," LangGraph when you need deterministic nodes mixed with agentic ones. Powerful. Easy to put the orchestrator in a system prompt and call it done.",[246,1643,1644,1647,1648,1650],{},[249,1645,1646],{},"MCP."," Plug. ",[193,1649,457],{"href":456},". Works behind frameworks and products. Does not choose the framework\u002Fharness cut.",[189,1652,1653,1654,1657,1658,1660,1661,219],{},"In Nimbus you do not import a graph to start a job. You assign an ",[193,1655,1656],{"href":679},"agent team"," on a ",[193,1659,1137],{"href":320},". Under the hood there is still a loop, tools, and stops — a harness. The product choice is whether operators must be graph authors. ",[193,1662,1663],{"href":482},"Self-service vs forward-deployed",[238,1665,1079],{"id":1078},[189,1667,1668,1669,1671],{},"Engineers will prefer frameworks. They should. Control, portability, tests in CI. Operators and Legal will prefer a harness they can inspect without a pull request. ",[193,1670,641],{"href":224}," keeps showing isolated technical use without operating-model change. A beautiful LangGraph in a platform team’s repo is still isolated use if RevOps cannot attach Salesforce or refuse a write.",[189,1673,1085],{},[243,1675,1676,1679,1682,1685],{},[246,1677,1678],{},"the RFP says “must support LangChain” as if that were a control",[246,1680,1681],{},"a vendor says “model-agnostic framework” and prices seats on one flagship",[246,1683,1684],{},"you are asked to rebuild quoting and SoD because “we already have agents in Python”",[246,1686,1687],{},"security reviews the GitHub org and never reviews who can call PATCH",[189,1689,1690,1693,1694,1697,1698,1701],{},[193,1691,1692],{"href":366},"OWASP’s LLM Top 10"," excessive agency shows up in both: a framework that exposes every tool by default, or a product that does. The cut is not safety vs convenience. It is ",[211,1695,1696],{},"who can change the harness when it fails"," — ",[193,1699,1700],{"href":477},"harness engineering"," — and whether a fail-closed write exists.",[189,1703,1704,1706],{},[193,1705,1144],{"href":1143}," Map\u002FMeasure need a system boundary. “Our framework” is not a boundary. A named runtime with grants and logs is.",[238,1708,1710],{"id":1709},"the-practical-differences","The practical differences",[189,1712,1713,1716],{},[249,1714,1715],{},"Who authors the loop."," Framework: software engineers. Product harness: operators (and maybe SE for custom tools). If only engineers can add a sensor, you will wait on a sprint for a Legal rule.",[189,1718,1719,1722,1723,1726,1727,1730],{},[249,1720,1721],{},"Where identity lives."," Framework default: service account in ",[390,1724,1725],{},".env",". Product harness: org roster, workstream membership, OAuth grants. You ",[211,1728,1729],{},"can"," do the latter in LangGraph. You must build it.",[189,1732,1733,1736,1737,322,1740,1742,1743,1745],{},[249,1734,1735],{},"What “done” means."," Framework: your node returned. Inner product harness: tests \u002F hook. Outer product harness: signer. Anthropic’s ",[193,1738,1739],{"href":373},"effective agents",[193,1741,467],{"href":466}," notes are about encoding done in the ",[211,1744,648],{},". Frameworks give you the primitives; they do not know your done.",[189,1747,1748,1751,1752,1755,1756,1758],{},[249,1749,1750],{},"Portability."," Frameworks win on model swap ",[211,1753,1754],{},"if"," tools and middleware stay. Product harnesses win if they actually route and do not bury a flagship default in a seat. ",[193,1757,314],{"href":313},". “We wrap LangChain” is not routing.",[189,1760,1761,1763],{},[249,1762,1221],{}," Frameworks shine in unit tests of nodes. Inner harnesses shine on SWE-bench \u002F Terminal-Bench. Enterprise harnesses shine when quote hash equals SoR row. Different CI.",[189,1765,1766,1769,1770,219],{},[249,1767,1768],{},"Time-to-first-governed-write."," Framework: months unless you already built the interceptor. Forward-deployed OS: months of people. Self-service outer harness: the product’s week-one claim — verify it. ",[193,1771,1773],{"href":1772},"\u002Fblog\u002Fhow-to-evaluate-an-agent-harness\u002F","Proof of value",[189,1775,1776,1779,1780,1783],{},[249,1777,1778],{},"Lock-in."," Framework lock-in is code and patterns. Product lock-in is data, graph, and operating habits. Both are real. ",[193,1781,1782],{"href":313},"How to solve model lock-in"," is the model slice; harness lock-in is the loop slice. Prefer quoted payloads and exportable ledgers either way.",[189,1785,1786,1787,1789,1790,1793,1794,219],{},"LangChain is not the villain. Their ",[193,1788,637],{"href":636}," is one of the clearer public derivations of harness parts. Use it. Then ask whether your ",[211,1791,1792],{},"deployment"," has those parts for the job you are buying — repo or company. ",[193,1795,1796],{"href":217},"Inner vs outer",[189,1798,1799,1800,1803],{},"Nimbus’s bet is that most operators should not author LangGraph to update a discount cap. The wiki and the gate should move. Teams that ",[211,1801,1802],{},"should"," author graphs (unique simulation, exotic tools) can still sit behind a connector. Framework inside a harness. Not a framework instead of one.",[238,1805,1807],{"id":1806},"a-decision-rule","A decision rule",[243,1809,1810,1816,1822,1830],{},[246,1811,1812,1815],{},[249,1813,1814],{},"Building a product or a unique workflow in code, with engineers on the hook:"," framework (or SDK harness) plus your own grants and evals.",[246,1817,1818,1821],{},[249,1819,1820],{},"Hiring a loop for a repository:"," inner product harness (Claude Code, Cursor, Codex). Optionally extend with a framework for custom tools.",[246,1823,1824,474,1827,1829],{},[249,1825,1826],{},"Hiring a loop for CRM\u002FERP\u002Fcross-department work:",[193,1828,1017],{"href":669}," \u002F OS-class product. A framework is a build programme.",[246,1831,1832,1835],{},[249,1833,1834],{},"Vendor says “we are a framework and an OS”:"," make them show a failed unsigned write and an operator-attached connector. Words are cheap.",[189,1837,1838,1841],{},[193,1839,1840],{"href":482},"Build vs buy an enterprise AI OS"," is the longer form of the third bullet.",[238,1843,1845],{"id":1844},"what-each-layer-of-the-stack-is-for","What each layer of the stack is for",[189,1847,1848,1849,1852,1853,1855],{},"LangChain’s own split is the cleanest vendor-native map: use Deep Agents when you want a batteries-included ",[211,1850,1851],{},"harness","; use ",[390,1854,1565],{}," when you want a minimal harness you customise with middleware; drop to LangGraph when the agent loop is the wrong shape and you need deterministic nodes mixed with agentic ones; use LangSmith to trace whatever you built. That is a builder’s menu. It does not decide whether RevOps can refuse a write.",[189,1857,1858,1859,1862,1863,1865],{},"CrewAI is a role-and-task framework. AutoGen is a conversation-of-agents framework. Semantic Kernel is Microsoft’s orchestration SDK. Pydantic AI moved toward a “harness-first” design in 2026 (capabilities as tools + hooks + instructions). None of these are wrong. All of them leave identity, SoR quoting, and operator self-service as ",[211,1860,1861],{},"your"," story unless you add them. ",[193,1864,367],{"href":366}," will still fail you if the first graph you merge attaches every production tool “so the demo looks alive.”",[189,1867,1868,1869,1871],{},"Product harnesses fail the other way: they hide the graph so operators can work, then surprise engineers who wanted to unit-test a node. Demand an escape hatch — export traces, typed payloads, maybe a documented tool SDK — without requiring every discount cap to be a pull request. Nimbus’s bet is that the cap lives in the ",[193,1870,675],{"href":674}," and the interceptor, and that engineers who need a custom simulator put it behind a connector. Framework inside the harness.",[189,1873,1874,1876],{},[193,1875,272],{"href":1339}," would say a company that standardises on LangGraph has invested in layer 2 (builder) and still has to build layers 3–4 (user guides\u002Fsensors, organisational ownership). A company that buys only a coding harness has a strong inner layer 2–3 and a missing outer layer 4. A company that buys an OS-class product is hoping layer 3–4 shipped. Verify with a refused write, not with a README.",[189,1878,1879,1882,1883,1885],{},[249,1880,1881],{},"Cost of the wrong cut."," Framework-first for operators: six months of platform work, then shadow copilots anyway. Product-first for a unique research loop: you will fight the product and rebuild the graph in Python by week four. ",[193,1884,493],{"href":492}," plus this page: workspace first, then assemble vs hire.",[189,1887,1888,1891,1892,1894,1895,1897,1898,1901],{},[249,1889,1890],{},"Portability, honestly."," Frameworks make model swap easier ",[211,1893,1754],{}," you used their model interface and did not sprinkle vendor-specific tool formats through application code. Products make operator ratchet easier ",[211,1896,1754],{}," adding a gate is a UI action. Neither gives you portability of ",[211,1899,1900],{},"decisions"," unless the ledger exports. Ask for JSON of the quote and the graph, not a promise of “open.”",[189,1903,1904,1905,1908],{},"Inngest and others have argued that durable execution needs “a harness, not a framework”: retries, state, and recovery as infrastructure. That slogan is directionally right for production. It is incomplete for enterprises. Durable retries of an ",[211,1906,1907],{},"unsigned"," write are a reliable incident. The outer harness adds identity and a stop that retries must not bypass. LangGraph checkpointing is excellent loop infrastructure. It is not a Finance signer.",[189,1910,1911,1912,219],{},"A worked split: the data-science team builds a forecasting graph in LangGraph, evaluates it with their own sensors, exposes it as a tool. RevOps never opens the repo. They brief a workstream, the team calls the forecast tool under read scope, and any CRM write still quotes in the product interceptor. Framework for the specialist. Harness for the company job. Nimbus is the second box; it should consume the first as a connector, not replace the scientists’ graph. ",[193,1913,50],{"href":451},[189,1915,1916],{},"If your platform team’s OKR is “stand up LangChain,” add a second OKR: “unsigned SoR writes are impossible.” The first without the second is a framework programme. The second without any loop is a policy PDF. You need both, in that order of safety.",[189,1918,1919,1920,1922],{},"CrewAI marketing will talk about roles. Roles in a YAML file are not roster identity. If the “legal reviewer” crew member can still call the same Salesforce write tool as the “AE,” you have a framework demo of ",[193,1921,1249],{"href":751}," without the contract. Ask to see the tool belt per role, then ask what happens when you remove the write tool from legal and the model asks for it anyway. The harness answer is refuse. The framework-only answer is often “we’ll prompt it.”",[238,1924,693],{"id":692},[507,1926,1928],{"id":1927},"is-langgraph-a-harness","Is LangGraph a harness?",[189,1930,1931,1932,1935],{},"It is a framework for building one. Your graph ",[211,1933,1934],{},"becomes"," a harness when it owns tool dispatch, bounds, and (for production) identity and sensors. Empty graph ≠ harness.",[507,1937,1939],{"id":1938},"is-claude-code-a-framework","Is Claude Code a framework?",[189,1941,1942,1943,219],{},"No. It is a productised inner harness. The Agent SDK is the embeddable form — closer to HaaS in ",[193,1944,1945],{"href":1446},"Osmani’s sense",[507,1947,1949],{"id":1948},"does-mcp-replace-both","Does MCP replace both?",[189,1951,1952],{},"No. Plumbing. Hosts still need a loop and grants.",[507,1954,1956],{"id":1955},"we-already-standardised-on-crewai","We already standardised on CrewAI.",[189,1958,1959],{},"Keep it for the jobs engineers should own. Do not force RevOps to write crews for a renewal write. Put CrewAI behind a scoped tool if the outer harness needs that specialist.",[507,1961,1963],{"id":1962},"how-do-we-evaluate-a-vendor-who-wraps-langchain","How do we evaluate a vendor who wraps LangChain?",[189,1965,1966,1967,1970],{},"Ignore the wrapper. Run ",[193,1968,1969],{"href":1772},"how to evaluate an agent harness",". If they cannot refuse a write, you evaluated a demo of a framework.",[507,1972,1974],{"id":1973},"where-does-nimbus-sit","Where does Nimbus sit?",[189,1976,1977,1978,1980,1981,219],{},"Productised outer harness, not a LangChain distribution. ",[193,1979,11],{"href":558},". You should still allow inner harnesses for code. ",[193,1982,493],{"href":492},[238,1984,746],{"id":745},[189,1986,1987,322,1989,219],{},[193,1988,741],{"href":477},[193,1990,1991],{"href":751},"what is multi-agent AI",[238,1993,760],{"id":759},[243,1995,1996,2000,2005,2009,2014,2019,2023,2027,2031,2035,2039,2043,2047,2051],{},[246,1997,1998],{},[193,1999,767],{"href":202},[246,2001,2002],{},[193,2003,2004],{"href":1585},"LangChain, How to build a custom agent harness",[246,2006,2007],{},[193,2008,772],{"href":636},[246,2010,2011],{},[193,2012,2013],{"href":1574},"LangChain, LangGraph overview",[246,2015,2016],{},[193,2017,2018],{"href":1569},"LangChain Deep Agents",[246,2020,2021],{},[193,2022,1452],{"href":1339},[246,2024,2025],{},[193,2026,1447],{"href":1446},[246,2028,2029],{},[193,2030,787],{"href":373},[246,2032,2033],{},[193,2034,792],{"href":466},[246,2036,2037],{},[193,2038,797],{"href":282},[246,2040,2041],{},[193,2042,802],{"href":224},[246,2044,2045],{},[193,2046,1144],{"href":1143},[246,2048,2049],{},[193,2050,816],{"href":366},[246,2052,2053],{},[193,2054,837],{"href":836},{"title":169,"searchDepth":170,"depth":170,"links":2056},[2057,2058,2059,2060,2061,2062,2070,2071],{"id":240,"depth":170,"text":241},{"id":1078,"depth":170,"text":1079},{"id":1709,"depth":170,"text":1710},{"id":1806,"depth":170,"text":1807},{"id":1844,"depth":170,"text":1845},{"id":692,"depth":170,"text":693,"children":2063},[2064,2065,2066,2067,2068,2069],{"id":1927,"depth":845,"text":1928},{"id":1938,"depth":845,"text":1939},{"id":1948,"depth":845,"text":1949},{"id":1955,"depth":845,"text":1956},{"id":1962,"depth":845,"text":1963},{"id":1973,"depth":845,"text":1974},{"id":745,"depth":170,"text":746},{"id":759,"depth":170,"text":760},"An agent framework is a library for assembling a loop.",{"eyebrow":861,"title":2074},"Library versus a system you can run",[2076,2079,2082],{"question":2077,"answer":2078},"Is an agent framework the same as an agent harness?","No. A framework is a library for assembling a loop. A harness is the loop you can actually run, with tools, stops, identity, and sensors included.",{"question":2080,"answer":2081},"Does LangChain count as a harness?","LangChain helps you build one. It is not, by itself, a harness you can hire. You still have to add permissions, a stop, and a record of what ran.",{"question":2083,"answer":2084},"What should we buy if we already use a framework?","Keep the framework for construction. Add a harness where the work touches a live system and a named person has to be able to refuse the change.","\u002Fblog\u002Fagent-harness-vs-agent-framework",{"title":1539,"description":2072},"blog\u002Fagent-harness-vs-agent-framework",[1531,878,2089,2090],"langchain","frameworks","RIUbWALaQhkECWWg7MZzfWGEoEgdzjiFlnkixJN_2ok",{"enabled":163,"message":2093,"linkLabel":79,"linkHref":80,"id":2094,"title":2095,"archived":163,"authors":164,"badge":164,"body":2096,"date":164,"definedTerm":164,"department":164,"description":169,"extension":172,"eyebrow":164,"faqHeader":164,"faqs":164,"footerBand":164,"headline":164,"image":164,"industry":164,"jobType":164,"listed":131,"location":164,"navigation":131,"openRoles":164,"pageLayout":164,"path":2100,"relatedHeading":164,"seo":2101,"series":164,"sitemap":163,"status":164,"stem":2102,"subhead":164,"tags":164,"video":164,"whyJoin":164,"workplaceType":164,"__hash__":2103},"We're hiring! Join the team building the Sentient Enterprise.","content\u002Fshared\u002Fhiring.md","Hiring banner",{"type":166,"value":2097,"toc":2098},[],{"title":169,"searchDepth":170,"depth":170,"links":2099},[],"\u002Fshared\u002Fhiring",{"title":2095,"description":169},"shared\u002Fhiring","1zs3boivKda1e-b-hAyuNcmZSKjZUAXmecnwHVgcHzk",{"fold":2105,"id":2109,"title":2110,"archived":163,"authors":164,"badge":164,"body":2111,"date":164,"definedTerm":164,"department":164,"description":169,"extension":172,"eyebrow":164,"faqHeader":164,"faqs":164,"footerBand":2115,"headline":164,"image":164,"industry":164,"jobType":164,"listed":131,"location":164,"navigation":131,"openRoles":164,"pageLayout":164,"path":2119,"relatedHeading":164,"seo":2120,"series":164,"sitemap":163,"status":164,"stem":2121,"subhead":164,"tags":164,"video":164,"whyJoin":164,"workplaceType":164,"__hash__":2122},{"headline":2106,"description":2107,"primaryLabel":8,"primaryTo":2108,"secondaryLabel":897,"secondaryTo":12},"Run frontier AI your business actually owns.","Governed workstreams, 3,000+ integrations, and a proprietary knowledge graph. Start on Free.","\u002Fsignup?plan=free","content\u002Fshared\u002Fcta.md","Site CTAs",{"type":166,"value":2112,"toc":2113},[],{"title":169,"searchDepth":170,"depth":170,"links":2114},[],{"headline":2116,"description":2117,"primaryLabel":8,"primaryTo":2108,"secondaryLabel":2118,"secondaryTo":85},"See what governed AI looks like on your stack.","Connect your tools, run a workstream, and keep every decision on your ledger. Start on Free.","Talk to our team","\u002Fshared\u002Fcta",{"title":2110,"description":169},"shared\u002Fcta","eYqahyaPnbp8GKrWpoORbZdtmkWmgHr5F61ZHOnb8sY",1792999485487]