[{"data":1,"prerenderedAt":698},["ShallowReactive",2],{"blog-cursor-en":3},{"slug":4,"published":5,"minutes":6,"category":7,"tags":8,"keywords":14,"about":21,"sources":25,"cover":47,"og":48,"expertise":49,"locales":50,"lang":51,"title":54,"description":55,"coverAlt":56,"metaTitle":57,"takeaways":58,"faq":64,"toc":77,"blocks":102,"others":429},"cursor","2026-09-02",10,"agents",[9,10,11,12,13],"AI code editor","Coding agents","CLI","Usage-based pricing","MCP",[4,15,16,17,18,19,20],"cursor vs github copilot","cursor pricing","cursor cli","ai code editor","cursor usage limits","cursor pro plus",[22],{"name":23,"url":24},"Cursor","https:\u002F\u002Fcursor.com",[26,29,32,35,38,41,44],{"title":27,"url":28},"Cursor pricing","https:\u002F\u002Fcursor.com\u002Fpricing",{"title":30,"url":31},"Cursor models and pricing docs","https:\u002F\u002Fcursor.com\u002Fdocs\u002Fmodels-and-pricing",{"title":33,"url":34},"Cursor CLI documentation","https:\u002F\u002Fcursor.com\u002Fdocs\u002Fcli\u002Foverview",{"title":36,"url":37},"GitHub Copilot plans","https:\u002F\u002Fgithub.com\u002Ffeatures\u002Fcopilot\u002Fplans",{"title":39,"url":40},"Claude pricing","https:\u002F\u002Fclaude.com\u002Fpricing",{"title":42,"url":43},"Claude Code cost documentation","https:\u002F\u002Fdocs.claude.com\u002Fen\u002Fdocs\u002Fclaude-code\u002Fcosts",{"title":45,"url":46},"Cline pricing","https:\u002F\u002Fcline.bot\u002Fpricing","\u002Fimages\u002Fblog\u002Fcursor\u002Fcover.webp","\u002Fimages\u002Fblog\u002Fcursor\u002Fog.jpg","ai-engineer",[51,52,53],"en","de","hu","Cursor reviewed: an AI code editor billed past its sticker price","Cursor bundles an editor, a terminal agent and cloud runs behind one subscription. What the two usage pools really cost, and when Copilot, Claude Code or Cline is the better buy.","Three rectangles in a row beneath one wide bracket, with two narrow columns hanging under it.","Cursor review: AI code editor and agent CLI · Balázs Csorba",[59,60,61,62,63],"Cursor's $20 Pro is the entrance price: the vendor's own docs put daily agent use at $60 to $100 a month of total usage, and power users at $200 or more.","Usage splits into two pools, Cursor Models for Grok and Composer at an included rate, and Other Models at each third-party model's API price.","On Teams and Enterprise every third-party request adds a Cursor Token Rate of $0.25 per million tokens, while first-party Grok and Composer are exempt.","No page publishes how many requests or tokens a Pro, Pro Plus or Ultra allowance contains, only that on-demand usage continues at the same API rates.","Privacy mode is what carries the guarantee that code is not used for training, so the promise depends on a setting rather than on the plan tier.",[65,68,71,74],{"q":66,"a":67},"How much does Cursor actually cost per month?","Pro is $20, Pro Plus $60 and Ultra $200, each covering two usage pools. Cursor's own documentation puts daily agent users at $60 to $100 of total usage a month and power users at $200 or more, and on-demand usage then bills in arrears at the same per-token API rates.",{"q":69,"a":70},"Cursor or GitHub Copilot for a small team?","Copilot Pro is $10 a month with unlimited completions and AI Credits at $0.01 each, so it is the cheaper seat. Cursor costs $20 and up but adds a fork editor, terminal CLI, cloud agents and review in one licence, which is only worth paying for if the team uses more than the editor.",{"q":72,"a":73},"Can Cursor run unattended in a CI pipeline?","Yes. The CLI installs with one curl command and has a print mode: agent -p with a prompt, a model name and an output format runs the same agent non-interactively, and agent ls plus agent resume picks the conversation up afterwards.",{"q":75,"a":76},"Does Cursor train on my code?","The pricing page guarantees that code data is not used for training by Cursor or its model providers when privacy mode is enabled, which an individual switches on or a team admin enforces. Teams plans include team-wide privacy mode, and Enterprise adds audit-level controls.",[78,81,84,87,90,93,96,99],{"id":79,"title":80},"what-it-is","What it is",{"id":82,"title":83},"how-it-works","How it works",{"id":85,"title":86},"getting-started","Getting started",{"id":88,"title":89},"pricing","Pricing",{"id":91,"title":92},"privacy-and-lock-in","Privacy and lock-in",{"id":94,"title":95},"where-it-shingles","Where it shingles",{"id":97,"title":98},"verdict","Verdict",{"id":100,"title":101},"sources","Sources",[103,107,110,113,121,139,140,143,149,152,156,159,175,176,179,181,184,226,227,230,287,290,304,305,308,320,323,324,327,373,376,377,380,394,398,399],{"type":104,"content":105},"paragraph",[106],"Cursor is an AI code editor built on a VS Code fork, sold by Anysphere, and by now also a terminal agent, a set of cloud agents and a code review bot that all draw from the same subscription. The position taken here: it is the most complete single seat in this market, and the $20 price on the pricing page is an entrance fee rather than a working budget, because the metering that actually decides what a month costs sits one documentation page deeper.",{"type":104,"content":108},[109],"It sits where a developer already spends the day, which is the whole strategy. Where Claude Code starts in a terminal and Copilot starts inside GitHub, Cursor owns the editor, then extends the same agent into the CLI, into CI and into cloud runs, so a team buys one licence instead of gluing four tools together. What it competes with is not just other editors but the habit of paying for a model API directly and running an open-source agent against it.",{"type":111,"level":112,"id":79,"text":80},"heading",2,{"type":104,"content":114},[115,116,120],"One product line with several surfaces: the desktop editor, the ",{"tag":117,"children":118},"code",[119],"agent"," CLI, cloud agents that run away from the machine, Bugbot for pull request review, and automations. The model list behind it mixes first-party models with the frontier API catalogue, and the docs table currently runs to dozens of entries from Anthropic, OpenAI, Google, Moonshot, Meta, Z.ai and Cursor's own Grok and Composer lines.",{"type":122,"ordered":123,"items":124},"list",false,[125,127,129,131,133,135,137],[126],"Vendor: Anysphere, Inc., sold only directly through cursor.com, with SOC 2, ISO 27001, ISO 42001 and AIUC-1 listed on the pricing page.",[128],"Surfaces: desktop editor, terminal CLI, cloud agents, mobile app, Bugbot code review, and automations with shared team context.",[130],"Individual plans: Hobby free, Pro at $20 a month, Pro Plus at $60, Ultra at $200, plus a Start plan at 649 rupees a month for developers in India.",[132],"Team plans: Teams Standard at $40 per user a month and Premium at $120, where Premium carries five times the Standard agent limits; Enterprise is custom priced.",[134],"Usage: two pools per billing cycle, Cursor Models for Grok and Composer, and Other Models billed at each third-party model's API rate.",[136],"CLI: installed with one curl command, three modes — Agent, Plan and Ask — and a print mode for scripts and CI.",[138],"Extensions of the same agent: MCP servers, skills, hooks and rules, plus a marketplace for team-wide distributions.",{"type":111,"level":112,"id":82,"text":83},{"type":104,"content":141},[142],"Every request leaves through one agent loop that gathers context from the workspace, applies rules and MCP tool results, and asks a model for the next edit; the editor, the CLI and the cloud agent are front ends on that loop rather than separate products. When Auto is selected, the Cursor Router picks a model per request from the Cost, Balance or Intelligence mode, and the docs are explicit that every Auto mode bills at the list price of whichever model the request landed on.",{"type":144,"src":145,"alt":146,"wide":147,"caption":148},"image","\u002Fimages\u002Fblog\u002Fcursor\u002Fdiagram-7.webp","How a Cursor request reaches a model Four boxes run left to right: the repository, the Cursor agent in the editor or the CLI, the router, and the model that answers. Two panels sit below: the Cursor Models pool covering Grok and Composer at included rates, and the Other Models pool, which draws at the third-party API price. Three lines state that on-demand usage is billed in arrears at the same API rates, that third-party requests on Teams and Enterprise carry a token rate of 0.25 dollars per million, and that print mode runs the same agent in a CI job with a text output format.",true,"One agent loop behind the editor and the CLI, with billing decided by which pool the chosen model belongs to.",{"type":104,"content":150},[151],"That pooling is the design decision worth arguing about. Choosing a first-party model draws from the Cursor Models pool at an included rate; choosing Claude or GPT draws from Other Models at that model's API price, so the model picker is a cost control as much as a quality one, and the two pools reset together with the monthly billing cycle.",{"type":111,"level":153,"id":154,"text":155},3,"modes-and-routing","Modes, routing and the tab model",{"type":104,"content":157},[158],"The interactive surfaces share one mode system, and the docs are careful to separate what is free from what is metered: completions and tab predictions are not the unit of billing, agent work is.",{"type":122,"ordered":123,"items":160},[161,167,169,171,173],[162,163,166],"Agent mode has full tool access, Plan mode designs the change first, and Ask mode is read-only; the CLI exposes all three through slash commands and ",{"tag":117,"children":164},[165],"--mode",".",[168],"Auto has three settings — Cost, Balance and Intelligence — and bills at the list price of the model each request is routed to.",[170],"Fast variants are priced above the standard rate — double on the Grok lines — and long context is billed at double input on some models while several Claude and Gemini lines state no long-context surcharge.",[172],"On legacy request-based plans, Max Mode extends a context window at the model's API rate plus 20%.",[174],"Tab completions and next-edit suggestions are unlimited on every paid plan, so the meter only starts when an agent is thinking.",{"type":111,"level":112,"id":85,"text":86},{"type":104,"content":177},[178],"The CLI is the shortest path to judging the tool: one install command, then the same agent that runs in the editor, with a print mode that turns it into a scriptable step. The snippet below covers the interactive case, the non-interactive case and resuming a conversation, which is the shape a CI job or a hook needs.",{"type":117,"code":180},"# install the CLI on macOS, Linux or WSL\ncurl https:\u002F\u002Fcursor.com\u002Finstall -fsS | bash\n\n# interactive session in the current repository\nagent \"add retry logic to the HTTP client and cover it with tests\"\n\n# non-interactive run, for a script or a CI job\nagent -p \"summarise the last commit and flag risky changes\" \\\n  --model \"gpt-5\" \\\n  --output-format text\n\n# pick the conversation up later\nagent ls\nagent resume",{"type":104,"content":182},[183],"Note what the print run does not need: no editor, no GUI, just a repository and an account. That matters for the review, because it means Cursor can be evaluated against a terminal agent like Claude Code on equal terms rather than as a prettier editor with chat in it.",{"type":185,"variant":186,"title":187,"body":188},"callout","warn","Three things to settle before the trial ends",[189],[190,191,192,193,194,195,196,197,196,198,199,193,194,200,199,192,193,201,192,202,202,201,193,203,191,192,193,201,192,200,203,204,193,203,191,192,193,205,206,197,201,193,201,192,202,202,193,203,191,192,193,207,201,200,199,192,166,193,208,198,209,205,192,210,200,198,205,193,207,201,200,199,192,193,211,196,197,211,201,193,196,198,193,206,198,197,192,193,203,191,192,193,196,198,197,202,207,205,192,205,193,200,210,206,207,198,203,193,196,201,193,197,206,198,201,207,210,192,205,212,193,213,196,202,202,192,205,193,196,198,193,200,195,195,192,200,195,201,193,200,203,193,203,191,192,193,201,200,210,192,193,214,215,216,193,195,200,203,192,201,212,193,200,198,205,193,203,191,192,193,205,206,197,201,193,194,207,203,193,205,200,196,202,217,193,200,199,192,198,203,193,207,201,192,193,200,203,193,218,219,220,193,203,206,193,218,221,220,220,193,200,193,210,206,198,203,191,193,206,222,193,203,206,203,200,202,193,207,201,200,199,192,193,223,196,203,191,193,194,206,223,192,195,193,207,201,192,195,201,193,200,203,193,218,224,220,220,193,206,195,193,210,206,195,192,166,193,215,195,196,225,200,197,217,193,210,206,205,192,193,196,201,193,200,193,201,192,203,203,196,198,199,212,193,201,206,193,203,191,192,193,198,206,209,203,195,200,196,198,196,198,199,193,199,207,200,195,200,198,203,192,192,193,206,198,193,197,206,205,192,193,200,194,194,202,196,192,201,193,223,191,192,198,193,196,203,193,196,201,193,201,223,196,203,197,191,192,205,193,206,198,212,193,192,196,203,191,192,195,193,194,192,195,201,206,198,200,202,202,217,193,206,195,193,192,198,222,206,195,197,192,205,193,213,217,193,200,193,203,192,200,210,193,200,205,210,196,198,166],"T","h","e"," ","p","r","i","c","n","g","a","s","l","t",";","d","o","u","O","-","m","k",",","b","A","P","I","y","$","6","0","1","f","w","2","v",{"type":111,"level":112,"id":88,"text":89},{"type":104,"content":228},[229],"Five individual and team tiers, all public except Enterprise. The subscription buys included model usage in two pools, and everything past it continues at the same per-token rates rather than being cut off.",{"type":231,"head":232,"rows":241},"table",[233,235,237,239],[234],"Plan",[236],"Price",[238],"Model usage",[240],"What it adds",[242,251,260,269,278],[243,245,247,249],[244],"Hobby",[246],"Free",[248],"Limited agent requests, Composer access",[250],"No card required; enough to judge the editor",[252,254,256,258],[253],"Pro",[255],"$20 a month",[257],"Both pools, on-demand afterwards",[259],"Unlimited tab completions, Bugbot, cloud agents",[261,263,265,267],[262],"Pro Plus",[264],"$60 a month",[266],"Both pools, larger allowance",[268],"Recommended by the vendor for daily agent use",[270,272,274,276],[271],"Ultra",[273],"$200 a month",[275],"Both pools, largest allowance",[277],"Recommended for agents running in parallel",[279,281,283,285],[280],"Teams",[282],"$40 or $120 per user",[284],"Standard, or Premium with 5x agent limits",[286],"SAML\u002FOIDC SSO, team privacy mode, usage analytics",{"type":104,"content":288},[289],"The gap worth noticing is that no page states how many requests or tokens a Pro, Pro Plus or Ultra allowance contains; the docs only say that model choice changes how fast it is consumed. Cursor's own guidance is the substitute: daily agent users typically land at $60 to $100 a month in total usage, and power users at $200 or more, which puts the honest entry price for real agent work at three times the sticker Pro.",{"type":185,"variant":186,"title":291,"body":292},"What sits on top of the per-token rate",[293],[208,198,193,190,192,200,210,201,193,200,198,205,193,294,198,203,192,195,194,195,196,201,192,212,193,203,191,196,195,205,209,194,200,195,203,217,193,210,206,205,192,202,193,195,192,295,207,192,201,203,201,193,197,200,195,195,217,193,200,193,296,207,195,201,206,195,193,190,206,211,192,198,193,297,200,203,192,193,206,222,193,218,220,166,224,298,193,194,192,195,193,210,196,202,202,196,206,198,193,203,206,211,192,198,201,193,206,198,193,203,206,194,193,206,222,193,203,191,192,193,210,206,205,192,202,299,201,193,214,215,216,193,194,195,196,197,192,212,193,222,206,195,193,196,198,197,202,207,205,192,205,212,193,206,198,209,205,192,210,200,198,205,193,200,198,205,193,213,195,196,198,199,209,217,206,207,195,209,206,223,198,209,211,192,217,193,207,201,200,199,192,193,200,202,196,211,192,204,193,222,196,195,201,203,209,194,200,195,203,217,193,300,195,206,211,193,200,198,205,193,296,206,210,194,206,201,192,195,193,200,195,192,193,192,301,192,210,194,203,166,193,208,194,203,196,198,199,193,196,198,203,206,193,195,192,199,196,206,198,200,202,193,205,200,203,200,193,195,192,201,196,205,192,198,197,217,193,200,205,205,201,193,221,220,302,193,203,206,193,210,206,205,192,202,193,194,195,196,197,196,198,199,212,193,200,198,205,193,303,200,301,193,303,206,205,192,193,206,198,193,202,192,199,200,197,217,193,194,202,200,198,201,193,200,205,205,201,193,224,220,302,193,203,206,193,203,191,192,193,214,215,216,193,195,200,203,192,166],"E","q","C","R","5","'","G","x","%","M",{"type":111,"level":112,"id":91,"text":92},{"type":104,"content":306},[307],"Two questions decide whether a company can standardise on this: what happens to the code, and how hard it is to leave.",{"type":122,"ordered":123,"items":309},[310,312,314,316,318],[311],"Privacy mode, enabled per user or enforced by a team admin, is what carries the guarantee that code data is not used for training by Cursor or by its model providers.",[313],"Teams add team-wide privacy mode enforcement, SAML and OIDC single sign-on, usage analytics and audit-style controls on Enterprise.",[315],"Models are hosted by the provider, a trusted partner or Cursor itself; the docs point at a public sub-processor list rather than promising local inference.",[317],"The editor is closed source and the subscription is the only route — the vendor states it sells directly and does not authorise resellers.",[319],"What does travel between tools: MCP servers, skills, hooks and rules are ordinary configuration, while the agent loop, tab model and Composer outputs are not portable.",{"type":104,"content":321},[322],"The lock-in is therefore real but layered: configuration is portable, history and agent behaviour are not. A team that standardises on Cursor's rules and skills keeps its instructions if it leaves, and loses the accumulated conversations, the tab model tuned to its codebase and the cloud agent workflows.",{"type":111,"level":112,"id":94,"text":95},{"type":104,"content":325},[326],"The weaknesses are structural. The sticker price is a floor and the real one is unreproducible before purchase, because allowance sizes are unpublished and usage depends on model choice; a Claude-heavy month and a Composer-heavy month on the same plan are not the same bill. The two-pool design also steers economics, not just quality: third-party models cost API money while first-party Grok and Composer are exempt from the Teams token rate and cheaper in the Cursor Models pool, so the path of least resistance is the vendor's own models. And an agent this integrated is an agent this hard to swap out.",{"type":231,"head":328,"rows":337},[329,331,333,335],[330],"Tool",[332],"What it sells",[334],"Billing unit",[336],"Where it wins",[338,346,355,364],[339,340,342,344],[23],[341],"Editor fork plus CLI, cloud agents and code review",[343],"$20 to $200 a seat, plus model usage at API rates",[345],"One licence covering editor, terminal and cloud runs",[347,349,351,353],[348],"GitHub Copilot",[350],"Assistant inside GitHub with CLI and cloud agent",[352],"Pro $10, Pro Plus $39, Max $100; AI Credits at $0.01 each",[354],"Cheapest credible seat, with unlimited completions",[356,358,360,362],[357],"Claude Code",[359],"Terminal agent on a subscription or API tokens",[361],"Pro $20, Max from $100; enterprise deployments report $150 to $250 per developer a month",[363],"Long autonomous sessions without owning an editor",[365,367,369,371],[366],"Cline",[368],"Open-source extension and CLI, bring your own key",[370],"Free software; inference at cost or your own API key",[372],"No subscription, no lock-in, full control of spend",{"type":104,"content":374},[375],"Read against those, Cursor is the expensive middle that buys coordination: Copilot undercuts it on seat price, Claude Code reports higher per-developer spend in exchange for staying out of the editor, and Cline removes the subscription entirely while leaving the operator to assemble the same loop. The judgement is that Cursor's premium is justified only where a team actually uses the editor, the CLI and the cloud agent — pay for one surface and the other two are dead weight.",{"type":111,"level":112,"id":97,"text":98},{"type":104,"content":378},[379],"Buy it as a seat, not as a $20 experiment. The editor is the product; the metering underneath is what a team has to instrument, and the honest budget for daily agent work starts at the Pro Plus tier rather than Pro.",{"type":122,"ordered":147,"items":381},[382,384,386,388,390,392],[383],"Use it when the team wants one licence covering the editor, a terminal agent, cloud runs and pull request review instead of four separate contracts.",[385],"Use it when the codebase benefits from a tuned tab model and workspace rules that accumulate in one place rather than per tool.",[387],"Skip it when per-developer spend must be predictable; unpublished allowances plus API-priced third-party models make forecasting guesswork without dashboard discipline.",[389],"Skip it when the team already lives in the terminal and GitHub — a Claude Code or Copilot subscription costs less and keeps the editor choice open.",[391],"Choose Pro Plus rather than Pro for agent work; Pro is priced for tab and chat habits, and the vendor's own usage guidance puts daily agent use at $60 to $100 a month.",[393],"Enable privacy mode before the first commit if any of the repository is covered by a confidentiality obligation, because the no-training guarantee is attached to that setting.",{"type":395,"content":396},"quote",[397],"Cursor is priced like an editor and billed like a model API. Judge the seat on how many of its surfaces the team actually opens, and the bill on which pool each model lands in.",{"type":111,"level":112,"id":100,"text":101},{"type":122,"ordered":147,"items":400},[401,405,409,413,417,421,425],[402,404],{"tag":200,"href":28,"children":403},[27]," — plan tiers, the usage FAQ and the privacy mode guarantee",[406,408],{"tag":200,"href":31,"children":407},[30]," — the two usage pools, per-model rates, the Cursor Token Rate and the usage guidance",[410,412],{"tag":200,"href":34,"children":411},[33]," — install command, interactive and print modes, sessions and sandbox controls",[414,416],{"tag":200,"href":37,"children":415},[36]," — Free, Pro, Pro+ and Max prices and the AI Credits unit",[418,420],{"tag":200,"href":40,"children":419},[39]," — Pro and Max prices, Claude Code inclusion and Team seat prices",[422,424],{"tag":200,"href":43,"children":423},[42]," — the reported enterprise cost per developer and month",[426,428],{"tag":200,"href":46,"children":427},[45]," — the free open-source tier and bring-your-own-key billing",[430,488,560,630],{"slug":431,"published":432,"minutes":433,"category":7,"tags":434,"keywords":439,"about":450,"sources":460,"cover":482,"og":483,"expertise":49,"locales":484,"lang":51,"title":485,"description":486,"coverAlt":487},"agent-tool-design-descriptions-errors","2026-10-17",11,[435,436,437,13,438],"AI agents","Tool use","Function calling","Agent design",[440,441,442,443,444,445,446,447,448,449],"agent tool design","writing tool descriptions for llm","function calling best practices","tool use error handling","mcp tool design","tool calling is_error","how many tools should an agent have","tool search defer_loading","agent computer interface","llm tool schema design",[451,454,457],{"name":452,"url":453},"Intelligent agent","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FIntelligent_agent",{"name":455,"url":456},"Model Context Protocol","https:\u002F\u002Fmodelcontextprotocol.io\u002F",{"name":458,"url":459},"Poka-yoke","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FPoka-yoke",[461,464,467,470,473,476,479],{"title":462,"url":463},"Anthropic Engineering: Writing effective tools for agents, with agents (Sep 2025)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fwriting-tools-for-agents",{"title":465,"url":466},"Anthropic Engineering: Building effective agents (Dec 2024)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fbuilding-effective-agents",{"title":468,"url":469},"Claude Platform Docs: Define tools","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Ftool-use\u002Fdefine-tools",{"title":471,"url":472},"Claude Platform Docs: Handle tool calls","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Ftool-use\u002Fhandle-tool-calls",{"title":474,"url":475},"Claude Platform Docs: Tool search tool","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Ftool-use\u002Ftool-search-tool",{"title":477,"url":478},"Model Context Protocol specification 2025-11-25: Tools","https:\u002F\u002Fmodelcontextprotocol.io\u002Fspecification\u002F2025-11-25\u002Fserver\u002Ftools",{"title":480,"url":481},"OpenAI API Docs: Function calling","https:\u002F\u002Fdevelopers.openai.com\u002Fapi\u002Fdocs\u002Fguides\u002Ffunction-calling","\u002Fimages\u002Fblog\u002Fagent-tool-design-descriptions-errors\u002Fcover.webp","\u002Fimages\u002Fblog\u002Fagent-tool-design-descriptions-errors\u002Fog.jpg",[51,52,53],"Designing tools for AI agents: names, descriptions, schemas and error messages","How to design agent tools that get picked and called correctly: task-shaped tools, descriptions, schemas, compact responses and errors a model can act on.","Three rectangular slots in a base, with a block of matching width lined up over the middle slot and two wrong-width blocks beside it.",{"slug":489,"published":432,"minutes":433,"category":7,"tags":490,"keywords":495,"about":508,"sources":518,"cover":554,"og":555,"expertise":49,"locales":556,"lang":51,"title":557,"description":558,"coverAlt":559},"coding-agents-legacy-codebase-migration",[491,10,492,493,494],"Legacy migration","Strangler fig","Testing","Code review",[496,497,498,499,500,501,502,503,504,505,506,507],"legacy code migration ai","coding agent legacy codebase","strangler fig pattern ai","characterization tests legacy code","ai code migration review","agent assisted refactoring","legacy modernization coding agent","differential testing migration","shadow traffic migration","ai code review gate","test harness before migration","llm code migration",[509,512,515],{"name":510,"url":511},"Legacy code","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FLegacy_code",{"name":513,"url":514},"Characterization test","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FCharacterization_test",{"name":516,"url":517},"Strangler fig pattern","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FStrangler_fig_pattern",[519,522,524,527,530,533,536,539,542,545,548,551],{"title":520,"url":521},"Martin Fowler: Strangler Fig (2024)","https:\u002F\u002Fmartinfowler.com\u002Fbliki\u002FStranglerFigApplication.html",{"title":523,"url":514},"Wikipedia: Characterization test",{"title":525,"url":526},"Google Research: Accelerating code migrations with AI (2024)","https:\u002F\u002Fresearch.google\u002Fblog\u002Faccelerating-code-migrations-with-ai\u002F",{"title":528,"url":529},"Ziftci et al.: Migrating Code At Scale With LLMs At Google (arXiv 2504.09691)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.09691",{"title":531,"url":532},"Anthropic: Effective harnesses for long-running agents (2025)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Feffective-harnesses-for-long-running-agents",{"title":534,"url":535},"Anthropic: Redesigning source control at Anthropic (2026)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fredesigning-source-control",{"title":537,"url":538},"METR: Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity","https:\u002F\u002Fmetr.org\u002Fblog\u002F2025-07-10-early-2025-ai-experienced-os-dev-study\u002F",{"title":540,"url":541},"Chand et al.: Why Industry Migrates and When It Trusts LLMs (ESEM 2026)","https:\u002F\u002Fdrops.dagstuhl.de\u002Fentities\u002Fdocument\u002F10.4230\u002FLIPIcs.ESEM.2026.85",{"title":543,"url":544},"Li et al.: Environment-in-the-Loop: Rethinking Code Migration with LLM-based Agents (arXiv 2602.09944)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.09944",{"title":546,"url":547},"AWS: Amazon Q Developer just reached a $260 million dollar milestone (2024)","https:\u002F\u002Faws.amazon.com\u002Fblogs\u002Fdevops\u002Famazon-q-developer-just-reached-a-260-million-dollar-milestone\u002F",{"title":549,"url":550},"OpenAI: Why SWE-bench Verified no longer measures frontier coding capabilities (2026)","https:\u002F\u002Fopenai.com\u002Findex\u002Fwhy-we-no-longer-evaluate-swe-bench-verified\u002F",{"title":552,"url":553},"GitHub Docs: About GitHub Copilot cloud agent","https:\u002F\u002Fdocs.github.com\u002Fen\u002Fcopilot\u002Fconcepts\u002Fagents\u002Fcoding-agent\u002Fabout-coding-agent","\u002Fimages\u002Fblog\u002Fcoding-agents-legacy-codebase-migration\u002Fcover.webp","\u002Fimages\u002Fblog\u002Fcoding-agents-legacy-codebase-migration\u002Fog.jpg",[51,52,53],"Migrating a legacy codebase with coding agents: a strangler fig playbook","How to use coding agents on a legacy migration: characterization tests first, one seam per task, review gates and differential verification.","A stack of slabs with the top two replaced by fresh ones beside a seam and a check.",{"slug":561,"published":432,"minutes":562,"category":7,"tags":563,"keywords":567,"about":579,"sources":589,"cover":624,"og":625,"expertise":49,"locales":626,"lang":51,"title":627,"description":628,"coverAlt":629},"background-agents-async-pr-workflow",12,[10,564,565,494,566],"Pull requests","CI","Automation",[568,569,570,571,572,573,574,575,576,577,578],"background coding agent","async coding agent","cloud coding agent pull request","copilot coding agent limits","codex cloud tasks","claude code cloud sessions","jules coding agent limits","agent pull request review","coding agent sandbox","agent CI gating","cost per merged pull request",[580,583,586],{"name":581,"url":582},"Software agent","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FSoftware_agent",{"name":584,"url":585},"Pull request","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FDistributed_version_control",{"name":587,"url":588},"Continuous integration","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FContinuous_integration",[590,591,594,597,600,603,606,609,612,615,618,621],{"title":552,"url":553},{"title":592,"url":593},"GitHub Docs: Risks and mitigations for Copilot cloud agent","https:\u002F\u002Fdocs.github.com\u002Fen\u002Fcopilot\u002Fconcepts\u002Fagents\u002Fcloud-agent\u002Frisks-and-mitigations",{"title":595,"url":596},"OpenAI Developers: Codex cloud environments","https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcloud\u002Fenvironments",{"title":598,"url":599},"Anthropic: Use Claude Code in the cloud","https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fclaude-code-on-the-web",{"title":601,"url":602},"Google Jules: Limits and plans","https:\u002F\u002Fjules.google\u002Fdocs\u002Fusage-limits\u002F",{"title":604,"url":605},"Google Jules: Getting started","https:\u002F\u002Fjules.google\u002Fdocs\u002F",{"title":607,"url":608},"GitHub Docs: About protected branches","https:\u002F\u002Fdocs.github.com\u002Fen\u002Frepositories\u002Fconfiguring-branches-and-merges-in-your-repository\u002Fmanaging-protected-branches\u002Fabout-protected-branches",{"title":610,"url":611},"GitHub Docs: Models and pricing for GitHub Copilot","https:\u002F\u002Fdocs.github.com\u002Fen\u002Fcopilot\u002Freference\u002Fcopilot-billing\u002Fmodels-and-pricing",{"title":613,"url":614},"OpenAI Developers: Codex pricing","https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fpricing",{"title":616,"url":617},"Ehsani et al.: Where Do AI Coding Agents Fail? An Empirical Study of Failed Agentic Pull Requests in GitHub (MSR 2026, arXiv 2601.15195)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2601.15195",{"title":619,"url":620},"Peralta et al.: Why Are Agentic Pull Requests Merged or Rejected? An Empirical Study (MSR 2026, arXiv 2605.22534)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.22534",{"title":622,"url":623},"DORA: State of AI-assisted Software Development 2025","https:\u002F\u002Fdora.dev\u002Fdora-report-2025","\u002Fimages\u002Fblog\u002Fbackground-agents-async-pr-workflow\u002Fcover.webp","\u002Fimages\u002Fblog\u002Fbackground-agents-async-pr-workflow\u002Fog.jpg",[51,52,53],"Background coding agents that open PRs: a practical async workflow","How async coding agents queue tasks, run in sandboxes, open draft PRs and pass CI. Provider limits, cost controls, review design and failure modes.","One card in a tray fanning out to five pads, all arrows leading to one page behind a gate.",{"slug":631,"published":632,"minutes":6,"category":7,"tags":633,"keywords":639,"about":651,"sources":659,"cover":692,"og":693,"expertise":49,"locales":694,"lang":51,"title":695,"description":696,"coverAlt":697},"context-rot-compaction-long-running-agents","2026-10-16",[634,635,636,637,638],"Context rot","Compaction","Context engineering","Long-running agents","Agent memory",[640,641,642,643,644,645,646,647,648,649,650],"context rot","context rot LLM","compaction AI agents","context engineering","long-running agents context window","Claude compaction API","context editing clear tool results","OpenAI responses compaction","lost in the middle LLM","agent context management","sub-agents context isolation",[652,655,658],{"name":653,"url":654},"Context window","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FLarge_language_model#Context_window",{"name":656,"url":657},"Attention (machine learning)","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FAttention_(machine_learning)",{"name":452,"url":453},[660,663,666,669,672,674,677,680,683,686,689],{"title":661,"url":662},"Chroma: Context Rot: How Increasing Input Tokens Impacts LLM Performance (Jul 2025)","https:\u002F\u002Fresearch.trychroma.com\u002Fcontext-rot",{"title":664,"url":665},"Liu et al.: Lost in the Middle: How Language Models Use Long Contexts (2023)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.03172",{"title":667,"url":668},"Modarressi et al.: NoLiMa: Long-Context Evaluation Beyond Literal Matching (2025)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.05167",{"title":670,"url":671},"Anthropic: Effective context engineering for AI agents (Sep 2025)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Feffective-context-engineering-for-ai-agents",{"title":673,"url":532},"Anthropic: Effective harnesses for long-running agents (Nov 2025)",{"title":675,"url":676},"Claude API docs: Compaction at a token threshold","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fbuild-with-claude\u002Fcompaction-threshold",{"title":678,"url":679},"Claude API docs: Compaction on demand","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fbuild-with-claude\u002Fcompaction-on-demand",{"title":681,"url":682},"Claude API docs: Context editing","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fbuild-with-claude\u002Fcontext-editing",{"title":684,"url":685},"Claude API docs: Memory tool","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Ftool-use\u002Fmemory-tool",{"title":687,"url":688},"OpenAI API docs: Compaction","https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fguides\u002Fcompaction",{"title":690,"url":691},"Cognition: Don't Build Multi-Agents","https:\u002F\u002Fcognition.ai\u002Fblog\u002Fdont-build-multi-agents","\u002Fimages\u002Fblog\u002Fcontext-rot-compaction-long-running-agents\u002Fcover.webp","\u002Fimages\u002Fblog\u002Fcontext-rot-compaction-long-running-agents\u002Fog.jpg",[51,52,53],"Context rot in long-running agents: clearing, compaction, memory and sub-agents","Why LLM accuracy drops as context grows, and how clearing tool results, compaction, notes and sub-agents keep long-running agents on track.","A small square press squeezing a tall stack of sheets into a short one, two notes branching off.",1792261155985]