[{"data":1,"prerenderedAt":628},["ShallowReactive",2],{"blog-cline-en":3},{"slug":4,"published":5,"minutes":6,"category":7,"tags":8,"keywords":14,"about":23,"sources":30,"cover":60,"og":61,"expertise":62,"locales":63,"lang":64,"title":67,"description":68,"coverAlt":69,"metaTitle":70,"takeaways":71,"faq":77,"toc":90,"blocks":115,"others":359},"cline","2026-06-16",11,"agents",[9,10,11,12,13],"Coding agent","Plan and Act","Checkpoints","MCP","Open source",[15,16,17,18,19,20,21,22],"cline coding agent","cline vs claude code","cline vs cursor","cline cli headless","coding agent auto approve","cline checkpoints","cline pass pricing","open source coding agent",[24,27],{"name":25,"url":26},"Cline","https:\u002F\u002Fgithub.com\u002Fcline\u002Fcline",{"name":28,"url":29},"Model Context Protocol","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FModel_Context_Protocol",[31,34,37,40,43,46,49,52,55,58],{"title":32,"url":33},"Cline docs: Overview","https:\u002F\u002Fdocs.cline.bot\u002Fintroduction\u002Foverview",{"title":35,"url":36},"Cline docs: Plan and Act mode","https:\u002F\u002Fdocs.cline.bot\u002Fcore-workflows\u002Fplan-and-act",{"title":38,"url":39},"Cline docs: CLI reference","https:\u002F\u002Fdocs.cline.bot\u002Fcli\u002Fcli-reference",{"title":41,"url":42},"Cline docs: Auto Approve and YOLO mode","https:\u002F\u002Fdocs.cline.bot\u002Ffeatures\u002Fauto-approve",{"title":44,"url":45},"Cline docs: Checkpoints","https:\u002F\u002Fdocs.cline.bot\u002Ffeatures\u002Fcheckpoints",{"title":47,"url":48},"Cline docs: MCP","https:\u002F\u002Fdocs.cline.bot\u002Fmcp\u002Fmcp-overview",{"title":50,"url":51},"Cline docs: Rules","https:\u002F\u002Fdocs.cline.bot\u002Fcustomization\u002Fcline-rules",{"title":53,"url":54},"Cline docs: ClinePass","https:\u002F\u002Fdocs.cline.bot\u002Fgetting-started\u002Fclinepass",{"title":56,"url":57},"Cline pricing","https:\u002F\u002Fcline.bot\u002Fpricing",{"title":59,"url":26},"cline\u002Fcline on GitHub","\u002Fimages\u002Fblog\u002Fcline\u002Fcover.webp","\u002Fimages\u002Fblog\u002Fcline\u002Fog.jpg","ai-engineer",[64,65,66],"en","de","hu","Cline, reviewed: an open coding agent that gives you every model and none of the guardrails","Cline is an Apache-2.0 coding agent for VS Code, JetBrains and the terminal. What Plan and Act, checkpoints and auto-approve actually guarantee, and where the safety model leaks.","Two tall panels side by side, a rail with three markers, and one square slipping out.","Cline, reviewed: an open coding agent · Balázs Csorba",[72,73,74,75,76],"Cline is Apache-2.0 and runs as a VS Code or JetBrains extension, a CLI, a desktop app and an SDK, all on the same agent core, so the approval model you learn in one place is the same in the others.","Plan mode cannot edit files or run commands by construction, which makes the plan a real artefact rather than a request the model is asked to honour.","Checkpoints commit to a separate shadow Git repository after every tool call, so rollback is cheap, at the cost of storing a full snapshot of the workspace per step on large repositories.","Auto-approve has no command allowlist: the model decides per call whether a command needs approval, and the CLI defaults to auto-approve true.","Because bring-your-own-key is the default, the security boundary is the provider, not the tool, and Cline adds no redaction, retention or audit layer of its own.",[78,81,84,87],{"q":79,"a":80},"Is Cline free?","The extension, CLI, desktop app and SDK are Apache-2.0 and free to install, with no seat fee. The model calls are not free unless you run a local model: Cline supports Claude, GPT, Gemini, Bedrock, Vertex, OpenRouter, Groq, Ollama and LM Studio, plus its own pay-as-you-go credits and ClinePass at $9.99 a month.",{"q":82,"a":83},"Is Cline safe to run with auto-approve on?","It is recoverable, not safe. Checkpoints snapshot the workspace after every tool call so you can roll the files back, but the agent still runs real commands on your machine while the snapshot is being written, and YOLO mode disables every check including file access outside the workspace. Run it on a branch, on a disposable machine, or inside a container.",{"q":85,"a":86},"Can Cline restrict which shell commands it runs?","Not with a built-in allow or deny list. The documentation states this plainly and points to a PreToolUse hook that cancels run_commands calls matching a pattern, which is the same mechanism used to enforce .clineignore now that .clineignore itself is being deprecated as a context filter.",{"q":88,"a":89},"Cline or Claude Code for an existing team?","Pick Cline when model choice is a policy requirement, when the bill must go to existing provider contracts, or when the team needs the SDK and CLI for automation. Pick Claude Code when you want one vendor's harness, one model's behaviour and one support path, and can accept the matching subscription.",[91,94,97,100,103,106,109,112],{"id":92,"title":93},"what-it-is","What Cline is",{"id":95,"title":96},"plan-and-act","Plan and Act, and why the split matters",{"id":98,"title":99},"getting-started","Getting started",{"id":101,"title":102},"pricing","What it costs",{"id":104,"title":105},"safety-model","The permission model, honestly",{"id":107,"title":108},"where-it-shingles","Where it shingle",{"id":110,"title":111},"verdict","Verdict",{"id":113,"title":114},"sources","Sources",[116,120,123,126,142,143,146,152,155,158,159,167,169,184,203,204,207,217,220,221,224,234,240,241,244,299,302,305,306,309,321,325,326],{"type":117,"content":118},"paragraph",[119],"Cline is a coding agent that runs where engineers already work, and it does not try to own the model. It ships as a VS Code and JetBrains extension, a CLI, a desktop app and an SDK, all built on one agent core published under Apache-2.0, and every one of them will connect to whatever provider the team has already bought. The position here is that this is the right default for anyone who has to justify the bill, and the wrong default for anyone who wants a guardrail they can rely on, because the permission model is softer than it looks.",{"type":117,"content":121},[122],"It competes directly with Claude Code, Cursor and the hosted coding assistants. Unlike those, it is not a product with a model attached but a harness with a model slot: the documentation lists Claude, GPT, Gemini, OpenRouter, AWS Bedrock, GCP Vertex, Groq, Cerebras, DeepSeek, local runtimes through Ollama and LM Studio, and anything speaking the OpenAI-compatible shape. That breadth is the feature, and it is also why comparing it to a closed product on quality is close to meaningless: most of what people praise or criticise about a coding agent is the model underneath.",{"type":124,"level":125,"id":92,"text":93},"heading",2,{"type":127,"ordered":128,"items":129},"list",false,[130,132,134,136,138,140],[131],"Apache-2.0, roughly 70,000 GitHub stars, 7,600 forks and 250-plus contributors, published as the cline CLI package and the @cline\u002Fsdk library.",[133],"The same agent core behind four surfaces: IDE extension, terminal CLI, desktop app for macOS, Windows and Linux, and a TypeScript SDK for embedding it elsewhere.",[135],"Plan mode and Act mode as separate states, with the option to run a different model in each.",[137],"Checkpoints on by default, committing a full snapshot of the workspace to a separate shadow Git repository after every tool call.",[139],"Per-category auto-approve toggles for reads, edits, commands, browser and MCP tools, plus a YOLO mode that turns all of them off.",[141],"MCP support with local STDIO servers and hosted Streamable HTTP endpoints, and rule files read from .clinerules, .cline\u002Frules, .cursorrules, .windsurfrules or AGENTS.md.",{"type":124,"level":125,"id":95,"text":96},{"type":117,"content":144},[145],"Plan mode is not a prompt. In that state the agent can read files, search and discuss, but it cannot modify a file or execute a command, and the conversation history carries over unchanged when Act mode starts. That makes the plan something the tool enforces rather than something the model is asked to remember, which is a materially better design than instructing a model to propose first.",{"type":147,"src":148,"alt":149,"wide":150,"caption":151},"image","\u002Fimages\u002Fblog\u002Fcline\u002Fdiagram-6.webp","The Cline Plan and Act loop with checkpoints A task enters Plan mode, where the agent reads files and searches but cannot write or run commands. Switching to Act mode carries the same conversation forward. Act mode then edits files and runs commands, and after every tool call a checkpoint is committed to a shadow Git repository. A restore returns the files, the task or both.",true,"Plan mode cannot write, so the plan is a constraint rather than a request. Act mode applies it, and every tool call is followed by a checkpoint in a separate Git repository that the project's own history never sees.",{"type":117,"content":153},[154],"Checkpoints are the other half of the design and deserve more attention than they get. After each file edit or command, Cline commits the current state of the working files to a shadow Git repository that is entirely separate from the project's real history. Three restore modes follow from that: restore the files and keep the conversation, restore the task and keep the code, or reset both. It is the reason auto-approve is defensible at all, and the documentation is candid that on very large repositories the snapshots cost storage and slow the agent down.",{"type":117,"content":156},[157],"Plan and Act can also point at different models, which is the cheapest cost lever in the whole product: a strong reasoning model for exploration and a cheap fast one for applying the plan. For small tasks the documentation recommends skipping planning entirely, because the planning pass is pure overhead when the answer is obvious.",{"type":124,"level":125,"id":98,"text":99},{"type":117,"content":160},[161,162,166],"The CLI is the surface worth learning first, because it is the only one where the approval defaults are visible and where the agent can run unattended. Headless mode switches on by itself when you pass ",{"tag":163,"children":164},"code",[165],"--json",", pipe stdin, or redirect stdout, which is a nice property for CI.",{"type":163,"code":168},"# Read every failing test and fix what caused them, then stop.\nset -euo pipefail\n\nnpm test 2>&1 | tee \u002Ftmp\u002Ffailures.log || true\n\n# Plan-only pass: no writes, no commands, just an analysis to review.\ncline -p \"Explain why these tests fail and what the smallest fix is\" \\\n  --json \u003C \u002Ftmp\u002Ffailures.log > \u002Ftmp\u002Fplan.jsonl\n\n# Apply pass on a clean branch, with a hard stop so a loop cannot burn the budget.\ncline \"Apply the fix for the failures listed in \u002Ftmp\u002Ffailures.log\" \\\n  --auto-approve true \\\n  --timeout 600 \\\n  --retries 3 \\\n  --model anthropic\u002Fclaude-sonnet-4-6 \\\n  | tee \u002Ftmp\u002Frun.log\n\n# Fail the job when the suite is still red, so the pipeline does not ship a broken tree.\nnpm test",{"type":117,"content":170},[171,172,175,176,179,180,183],"Three defaults in that invocation deserve scrutiny. ",{"tag":163,"children":173},[174],"--auto-approve"," defaults to true in the CLI, so unattended runs are the out-of-the-box behaviour, not the exception. ",{"tag":163,"children":177},[178],"--retries"," caps consecutive mistakes, which is the only guard against an agent looping on the same failing edit. And ",{"tag":163,"children":181},[182],"--timeout"," is the wall-clock stop that keeps a scheduled job from running until the budget is gone.",{"type":185,"variant":186,"title":187,"body":188},"callout","warn","The CLI has no command allowlist at all",[189],[190,191,194,195,198,199,202],"The documentation says so directly: to block shell commands you have to install a ",{"tag":163,"children":192},[193],"PreToolUse"," hook that cancels ",{"tag":163,"children":196},[197],"run_commands"," calls matching a pattern. The same mechanism is now the recommended way to enforce ",{"tag":163,"children":200},[201],".clineignore",", which is being deprecated precisely because it was never an access boundary: ignored files could still be read through an explicit mention or a shell command.",{"type":124,"level":125,"id":101,"text":102},{"type":117,"content":205},[206],"The software is free and the inference is not. The extension, CLI, desktop app and SDK carry no seat fee, so the marginal cost of running Cline is whatever the chosen provider charges for the tokens an agentic loop burns, and an agent loop burns many times more tokens than a chat because every tool result comes back into the context. There is no seat price to negotiate and no invoice from Cline for the open-source path.",{"type":127,"ordered":128,"items":208},[209,211,213,215],[210],"Bring your own key: pay the provider directly, under a contract and a spend limit the organisation already controls.",[212],"Cline usage-billing: one account, prepaid credits, one balance across the models Cline fronts, with some models tagged free for limited periods.",[214],"ClinePass at $9.99 a month, a flat subscription that the documentation advertises as two to five times the usage on a curated set of open coding models against standard API rates.",[216],"Enterprise: SSO, role-based access, centralised billing, audit logs, VPC deployment and OpenTelemetry, none of which exists in the free tier.",{"type":117,"content":218},[219],"The practical consequence is that cost control moves out of the tool and into the provider console. A team that wants a hard ceiling on agent spend needs either a provider that enforces one or a wrapper around the CLI that fails the run on a token or dollar budget, because Cline exposes usage in its session history and its enterprise dashboard but does not itself stop a run when a threshold is reached.",{"type":124,"level":125,"id":104,"text":105},{"type":117,"content":222},[223],"This is the part worth being blunt about. Auto-approve is evaluated per tool call against a category toggle, but there is no fixed allowlist of commands. The model marks each command as safe or requiring approval based on the command and its arguments, and the documentation gives examples rather than guarantees: build and test commands are usually safe, while installs, deletions, moves and in-place edits usually need approval. A security control that is implemented by asking a language model to classify a shell string is a speed bump, not a boundary.",{"type":127,"ordered":128,"items":225},[226,228,230,232],[227],"YOLO mode auto-approves everything: file operations anywhere on the machine, all terminal commands, browser actions, MCP tools and even the Plan to Act transition.",[229],"The read and edit toggles have an all-files variant, which extends access outside the workspace when the base toggle is on.",[231],"Scheduling, agent teams and subagents exist in the SDK, CLI and Kanban but not in the IDE extensions, so a control tested in one surface may not exist in another.",[233],"Subagents are read-only by construction, which is the one part of the tool where a capability limit is real rather than advisory.",{"type":185,"variant":235,"title":236,"body":237},"note","Run it where a mistake is cheap",[238],[239],"The combination that works is a throwaway branch or worktree, a container or a throwaway virtual machine, and a spend cap at the provider. Checkpoints then make the loop fast without making it safe, and the residual risk is whatever the agent ran before the snapshot caught up.",{"type":124,"level":125,"id":107,"text":108},{"type":117,"content":242},[243],"The fragmentation is the tax. Features land in the SDK and CLI first and reach the extensions later, which the documentation states outright for scheduling, agent teams and subagents. The JetBrains plugin is the sharpest version of the problem: the repository states plainly that JetBrains plugins are not being open-sourced, so the IDE that a large part of the European developer base runs on is the one surface a reviewer cannot audit.",{"type":245,"head":246,"rows":254},"table",[247,249,250,252],[248],"Dimension",[25],[251],"Claude Code",[253],"Cursor",[255,263,272,281,290],[256,258,260,262],[257],"Licence",[259],"Apache-2.0, source available",[261],"Proprietary",[261],[264,266,268,270],[265],"Model choice",[267],"Any provider, own key, local weights",[269],"Claude only",[271],"Several providers, own key",[273,275,277,279],[274],"Headless and CI",[276],"First-class CLI, SDK and cron scheduling",[278],"CLI, first-class",[280],"Cloud agents and CLI",[282,284,286,288],[283],"Command restriction",[285],"Hook-based, no built-in allowlist",[287],"Permission rules",[289],"Admin-controlled allow and deny",[291,293,295,297],[292],"Cost shape",[294],"Free software, provider bill",[296],"Subscription plus API",[298],"Subscription plus usage",{"type":117,"content":300},[301],"Context handling is the other soft spot, and it is not specific to Cline. An agent that reads files, runs tests, reads the failures and reads the diff again will spend most of its budget re-reading. The documented answer is subagents, which explore in parallel with their own context windows and return a short report, and memory bank files for structure. Both help, and neither changes the fact that an agentic session on a large repository is expensive in tokens relative to the value of the change.",{"type":117,"content":303},[304],"Finally, the tool's own security surface deserves the same scrutiny as any other agent: the repository has a security policy, and a project that lets an agent read files and run commands on a developer machine is a dependency with a shell. Treat the version as a pinned dependency, review the changelog before upgrades, and keep the provider keys scoped.",{"type":124,"level":125,"id":110,"text":111},{"type":117,"content":307},[308],"Cline is the best answer available for a team that has to choose its model, keep its data inside an existing provider contract, or automate agent work in CI without a vendor's permission system in the way. It is not a better version of a closed coding agent, and reviews that rank the two are comparing a harness to a model.",{"type":127,"ordered":150,"items":310},[311,313,315,317,319],[312],"Teams whose provider choice is a procurement decision rather than a preference.",[314],"Engineers automating repository work in CI, where the headless CLI and the SDK are the reason to adopt it.",[316],"Anyone building on top of an agent runtime, because @cline\u002Fsdk is the agent core rather than a wrapper.",[318],"Do not adopt it on the assumption that checkpoints make autonomy safe. They make it recoverable.",[320],"Do not adopt it where the command surface must be provably bounded without writing hooks, or where an unauditable IDE plugin disqualifies the tool.",{"type":185,"variant":235,"body":322},[323],[324],"The trade is explicit and worth stating plainly: you gain model choice, an auditable licence and real headless automation, and you pay for it with a permission model that leans on the model to classify commands and with features that arrive in the CLI before they arrive in the editor.",{"type":124,"level":125,"id":113,"text":114},{"type":127,"ordered":150,"items":327},[328,332,335,338,341,344,347,350,353,356],[329],{"tag":330,"href":33,"children":331},"a",[32],[333],{"tag":330,"href":36,"children":334},[35],[336],{"tag":330,"href":39,"children":337},[38],[339],{"tag":330,"href":42,"children":340},[41],[342],{"tag":330,"href":45,"children":343},[44],[345],{"tag":330,"href":48,"children":346},[47],[348],{"tag":330,"href":51,"children":349},[50],[351],{"tag":330,"href":54,"children":352},[53],[354],{"tag":330,"href":57,"children":355},[56],[357],{"tag":330,"href":26,"children":358},[59],[360,416,489,559],{"slug":361,"published":362,"minutes":6,"category":7,"tags":363,"keywords":368,"about":379,"sources":388,"cover":410,"og":411,"expertise":62,"locales":412,"lang":64,"title":413,"description":414,"coverAlt":415},"agent-tool-design-descriptions-errors","2026-10-17",[364,365,366,12,367],"AI agents","Tool use","Function calling","Agent design",[369,370,371,372,373,374,375,376,377,378],"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",[380,383,385],{"name":381,"url":382},"Intelligent agent","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FIntelligent_agent",{"name":28,"url":384},"https:\u002F\u002Fmodelcontextprotocol.io\u002F",{"name":386,"url":387},"Poka-yoke","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FPoka-yoke",[389,392,395,398,401,404,407],{"title":390,"url":391},"Anthropic Engineering: Writing effective tools for agents, with agents (Sep 2025)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fwriting-tools-for-agents",{"title":393,"url":394},"Anthropic Engineering: Building effective agents (Dec 2024)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fbuilding-effective-agents",{"title":396,"url":397},"Claude Platform Docs: Define tools","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Ftool-use\u002Fdefine-tools",{"title":399,"url":400},"Claude Platform Docs: Handle tool calls","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Ftool-use\u002Fhandle-tool-calls",{"title":402,"url":403},"Claude Platform Docs: Tool search tool","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Ftool-use\u002Ftool-search-tool",{"title":405,"url":406},"Model Context Protocol specification 2025-11-25: Tools","https:\u002F\u002Fmodelcontextprotocol.io\u002Fspecification\u002F2025-11-25\u002Fserver\u002Ftools",{"title":408,"url":409},"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",[64,65,66],"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":417,"published":362,"minutes":6,"category":7,"tags":418,"keywords":424,"about":437,"sources":447,"cover":483,"og":484,"expertise":62,"locales":485,"lang":64,"title":486,"description":487,"coverAlt":488},"coding-agents-legacy-codebase-migration",[419,420,421,422,423],"Legacy migration","Coding agents","Strangler fig","Testing","Code review",[425,426,427,428,429,430,431,432,433,434,435,436],"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",[438,441,444],{"name":439,"url":440},"Legacy code","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FLegacy_code",{"name":442,"url":443},"Characterization test","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FCharacterization_test",{"name":445,"url":446},"Strangler fig pattern","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FStrangler_fig_pattern",[448,451,453,456,459,462,465,468,471,474,477,480],{"title":449,"url":450},"Martin Fowler: Strangler Fig (2024)","https:\u002F\u002Fmartinfowler.com\u002Fbliki\u002FStranglerFigApplication.html",{"title":452,"url":443},"Wikipedia: Characterization test",{"title":454,"url":455},"Google Research: Accelerating code migrations with AI (2024)","https:\u002F\u002Fresearch.google\u002Fblog\u002Faccelerating-code-migrations-with-ai\u002F",{"title":457,"url":458},"Ziftci et al.: Migrating Code At Scale With LLMs At Google (arXiv 2504.09691)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2504.09691",{"title":460,"url":461},"Anthropic: Effective harnesses for long-running agents (2025)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Feffective-harnesses-for-long-running-agents",{"title":463,"url":464},"Anthropic: Redesigning source control at Anthropic (2026)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Fredesigning-source-control",{"title":466,"url":467},"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":469,"url":470},"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":472,"url":473},"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":475,"url":476},"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":478,"url":479},"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":481,"url":482},"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",[64,65,66],"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":490,"published":362,"minutes":491,"category":7,"tags":492,"keywords":496,"about":508,"sources":518,"cover":553,"og":554,"expertise":62,"locales":555,"lang":64,"title":556,"description":557,"coverAlt":558},"background-agents-async-pr-workflow",12,[420,493,494,423,495],"Pull requests","CI","Automation",[497,498,499,500,501,502,503,504,505,506,507],"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",[509,512,515],{"name":510,"url":511},"Software agent","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FSoftware_agent",{"name":513,"url":514},"Pull request","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FDistributed_version_control",{"name":516,"url":517},"Continuous integration","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FContinuous_integration",[519,520,523,526,529,532,535,538,541,544,547,550],{"title":481,"url":482},{"title":521,"url":522},"GitHub Docs: Risks and mitigations for Copilot cloud agent","https:\u002F\u002Fdocs.github.com\u002Fen\u002Fcopilot\u002Fconcepts\u002Fagents\u002Fcloud-agent\u002Frisks-and-mitigations",{"title":524,"url":525},"OpenAI Developers: Codex cloud environments","https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fcloud\u002Fenvironments",{"title":527,"url":528},"Anthropic: Use Claude Code in the cloud","https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fclaude-code-on-the-web",{"title":530,"url":531},"Google Jules: Limits and plans","https:\u002F\u002Fjules.google\u002Fdocs\u002Fusage-limits\u002F",{"title":533,"url":534},"Google Jules: Getting started","https:\u002F\u002Fjules.google\u002Fdocs\u002F",{"title":536,"url":537},"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":539,"url":540},"GitHub Docs: Models and pricing for GitHub Copilot","https:\u002F\u002Fdocs.github.com\u002Fen\u002Fcopilot\u002Freference\u002Fcopilot-billing\u002Fmodels-and-pricing",{"title":542,"url":543},"OpenAI Developers: Codex pricing","https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fpricing",{"title":545,"url":546},"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":548,"url":549},"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":551,"url":552},"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",[64,65,66],"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":560,"published":561,"minutes":562,"category":7,"tags":563,"keywords":569,"about":581,"sources":589,"cover":622,"og":623,"expertise":62,"locales":624,"lang":64,"title":625,"description":626,"coverAlt":627},"context-rot-compaction-long-running-agents","2026-10-16",10,[564,565,566,567,568],"Context rot","Compaction","Context engineering","Long-running agents","Agent memory",[570,571,572,573,574,575,576,577,578,579,580],"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",[582,585,588],{"name":583,"url":584},"Context window","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FLarge_language_model#Context_window",{"name":586,"url":587},"Attention (machine learning)","https:\u002F\u002Fen.wikipedia.org\u002Fwiki\u002FAttention_(machine_learning)",{"name":381,"url":382},[590,593,596,599,602,604,607,610,613,616,619],{"title":591,"url":592},"Chroma: Context Rot: How Increasing Input Tokens Impacts LLM Performance (Jul 2025)","https:\u002F\u002Fresearch.trychroma.com\u002Fcontext-rot",{"title":594,"url":595},"Liu et al.: Lost in the Middle: How Language Models Use Long Contexts (2023)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.03172",{"title":597,"url":598},"Modarressi et al.: NoLiMa: Long-Context Evaluation Beyond Literal Matching (2025)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.05167",{"title":600,"url":601},"Anthropic: Effective context engineering for AI agents (Sep 2025)","https:\u002F\u002Fwww.anthropic.com\u002Fengineering\u002Feffective-context-engineering-for-ai-agents",{"title":603,"url":461},"Anthropic: Effective harnesses for long-running agents (Nov 2025)",{"title":605,"url":606},"Claude API docs: Compaction at a token threshold","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fbuild-with-claude\u002Fcompaction-threshold",{"title":608,"url":609},"Claude API docs: Compaction on demand","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fbuild-with-claude\u002Fcompaction-on-demand",{"title":611,"url":612},"Claude API docs: Context editing","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fbuild-with-claude\u002Fcontext-editing",{"title":614,"url":615},"Claude API docs: Memory tool","https:\u002F\u002Fplatform.claude.com\u002Fdocs\u002Fen\u002Fagents-and-tools\u002Ftool-use\u002Fmemory-tool",{"title":617,"url":618},"OpenAI API docs: Compaction","https:\u002F\u002Fplatform.openai.com\u002Fdocs\u002Fguides\u002Fcompaction",{"title":620,"url":621},"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",[64,65,66],"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.",1792261156484]