Skip to content
Agent Friendly Code

AI contribution policies in 120 popular open source repos

120 of the 388 popular GitHub repos we scanned have a written AI contribution policy we could find. 9 ban AI-generated code, 40 restrict it, 31 accept it with disclosure, 34 allow it and 6 actively welcome it. Checked on 16 Nov 2026; a repo with no policy here may still have one somewhere we didn't look.

Written policy found

31%

120 of 388 GitHub repos

Require disclosure

46%

55 of 120 policies

Want PR text in your words

37%

44 policies rule out AI-written descriptions, issues or replies

Ban AI code outright

8%

9 of 120 policies

Stance breakdown

A policy that fits two stances takes the stricter one. Requiring human-written PR descriptions or replies counts as restricting, even where AI code is allowed.

What most policies ask for

  • Disclosure: 55 of 120 require you to say you used AI, in the PR description, a label or an Assisted-by commit trailer.
  • Your own words: 44 require PR descriptions, issues or replies to maintainers to be written by you; translation help is usually still fine.
  • A human in the loop: 33 explicitly refuse autonomous agents that open PRs or post comments on their own.

If you contribute with an AI agent

  1. Read the repo's CONTRIBUTING file and any AI_POLICY.md before the agent writes code. Many policies only accept AI work for issues a maintainer has accepted.
  2. Disclose the tool and how much it did, in the place the policy names.
  3. Understand every line, run the tests yourself, and answer review questions in your own words.
  4. Point the agent at the repo's AGENTS.md, so it follows the project's conventions from the start.

See also: projects easy for an agent to work in, how to set an agent up and what a CONTRIBUTING guide covers.

Projects people ask about

  • Rust Restricts LLM-created PRs need disclosure, a pre-agreed reviewer and non-critical scope; LLM-written comments and descriptions are banned.
  • Kubernetes: no AI policy in the repo's own contributing files when we checked.
  • Node.js Disclose AI allowed with disclosure and verification; no automated PRs, AI good-first-issue fixes or wholly AI-written messages.
  • LLVM Disclose Any tool with a human in the loop; label substantial AI content; no unapproved agents or good-first-issue fixes.
  • TypeScript Disclose Fully agent-written patches are fine if understood and disclosed; bulk autonomous agent PRs are never accepted.

Every policy, by stance

Jump to bans, restricts, disclose, allows, welcomes. Summaries are ours; the linked file is the authority.

Bans9 repos

AI-generated code or pull requests are not accepted.

Repos whose AI policy: AI-generated code or pull requests are not accepted.
RepoWhat it saysPolicy
alacritty/alacrittyNo content from language models in code, docs, issues, pull requests or comments.CONTRIBUTING.md
Anuken/MindustryNo AI-generated PRs or posted text; AI may review or translate, and AI-sourced findings must be marked.CONTRIBUTING.md
gleam-lang/gleamContributions from AI agents are not accepted; bug reports, PR descriptions and comments must be human-written.CONTRIBUTING.md
immich-app/immichDo not open LLM-generated PRs or post LLM-written descriptions or comments; translation is acceptable.CONTRIBUTING.md
lysine-dev/okhttpNo LLM-generated code, docs or communication; only disclosed translation and local autocomplete are allowed.CONTRIBUTING.md
lysine-dev/retrofitNo LLM-generated code, docs or communication; only disclosed translation and local autocomplete are allowed.CONTRIBUTING.md
meilisearch/meilisearchAI-generated content is not considered in external contributions, including code, tests, docs, PRs and issues.CONTRIBUTING.md
obsproject/obs-studioAll code, descriptions, issues and comments must be human-written; assistive AI such as Copilot is heavily discouraged.CONTRIBUTING.md
PostgREST/postgrestAdopts Gentoo's policy, which forbids contributing anything created with AI language tools.CONTRIBUTING.md

Restricts40 repos

AI code is accepted only in narrow ways: with prior approval, for some kinds of change, or with human-written PR text and replies.

Repos whose AI policy: AI code is accepted only in narrow ways: with prior approval, for some kinds of change, or with human-written PR text and replies.
RepoWhat it saysPolicy
Alamofire/AlamofireEvery PR states how much AI helped; PR descriptions must be human-written; AI translation is allowed.CONTRIBUTING.md
anomalyco/opencodeLong AI-generated PR descriptions and issues may be ignored; explain changes briefly in your own words.CONTRIBUTING.md
astral-sh/ruffFollows Astral's policy: AI coding allowed, but comments and PR text are human-written; no autonomous agents.CONTRIBUTING.md
astral-sh/uvFollows Astral's policy: AI coding allowed, but comments and PR text are human-written; no autonomous agents.CONTRIBUTING.md
bevyengine/bevyLinked policy tolerates careful AI coding with disclosure; AI must not write docs, issues or PR prose.CONTRIBUTING.md
biomejs/biomeDisclose any AI assistance and its extent; do not use AI to write PR descriptions or communication.CONTRIBUTING.md
BurntSushi/ripgrepAI coding is welcome, but comments and PR explanations must be human-written; autonomous agents are not allowed.AI_POLICY.md
clap-rs/clapAI coding needs maintainer permission on the issue first and disclosure; comments must be human; no autonomous agents.AI_POLICY.md
duckdb/duckdbAI coding is welcome, but comments and PR explanations must be human-written; autonomous agents are not allowed.AI_POLICY.md
elastic/elasticsearchDo not paste LLM output into PR descriptions or use LLM tools to talk with maintainers.CONTRIBUTING.md
elixir-lang/elixirAgents only for issues labelled Contributions Welcome or accepted features; disclose AI with an Assisted-by tag.CONTRIBUTING.md
eslint/eslintAI-generated PRs must be disclosed and target accepted issues; maintainers expect human replies to feedback.docs/src/contribute/ai-policy.md
ggml-org/llama.cppAI code allowed with disclosure and thorough manual review; AI must not write issues, PR text or replies.CONTRIBUTING.md
ggml-org/whisper.cppNo mostly AI-generated PRs; AI may only assist human-written code, with disclosure; no AI-written posts.CONTRIBUTING.md
gohugoio/hugoDisclose substantial AI use; AI contributions from non-maintainers must stay narrow, such as bug fixes.CONTRIBUTING.md
hashicorp/terraformDisclose the role AI played; PR descriptions must be human-crafted; no submissions from bot accounts..github/CONTRIBUTING.md
Kotlin/kotlinx.coroutinesLLM output only for trivial, single-obvious-way changes; non-trivial code and discussion must not be LLM-authored.CONTRIBUTING.md
ktorio/ktorAny tool with human review; label substantial generated content; never use AI for comments or reviews.AI_POLICY.md
localsend/localsendAI-generated contributions are refused unless they are bug fixes, very small, or from proven experts.CONTRIBUTING.md
mastodon/mastodonLinked policy discourages AI; disclose significant use; no AI-written PR descriptions or comments; no autonomous agents.CONTRIBUTING.md
medusajs/medusaAI may help write code, but PR descriptions and replies to reviewers must be in your words.CONTRIBUTING.md
n8n-io/n8nAI-assisted work welcome if you understand every line; write descriptions yourself; disclosure encouraged, not required.CONTRIBUTING.md
nats-io/nats-serverAI-assisted code accepted if understood and tested; communicate in your own words, not through an AI.CONTRIBUTING.md
nextcloud/serverDisclose AI with Assisted-by trailers; descriptions and replies in your own words; no autonomous agents..github/CONTRIBUTING.md
nuxt/nuxtAI welcome under two rules: write comments, issues and PR descriptions yourself, and understand what you submit.CONTRIBUTING.md
pola-rs/polarsNo agent interaction; disclose all AI use; AI code only for accepted issues and never good first issues.AI_POLICY.md
qdrant/qdrantWrite the PR motivation yourself; disclose AI-made commits and the initial prompt; you own the result.docs/CONTRIBUTING.md
rust-lang/cargoFollows Rust's LLM policy: disclosed, pre-arranged LLM-created PRs only; no LLM-written comments, issues or descriptions.CONTRIBUTING.md
rust-lang/rustLLM-created PRs need disclosure, a pre-agreed reviewer and non-critical scope; LLM-written comments and descriptions are banned.CONTRIBUTING.md
rust-lang/rustlingsLLM-generated code must be disclosed; using LLMs to write comments, issues or PR descriptions is banned.CONTRIBUTING.md
scala/scala3AI-assisted code accepted with full responsibility; all issues, PRs and discussions must be entirely human-written.LLM_POLICY.md
sharkdp/fdDisclose AI use and extent; review generated code yourself; PR descriptions and comments must be human-written.CONTRIBUTING.md
sqlalchemy/sqlalchemyUsing AI for help is fine, but don't post AI-generated text, even lightly edited; use your own words..github/CONTRIBUTING.md
TanStack/queryAI may generate code you fully understand; comments, issues and PR descriptions must be in your voice.AI_POLICY.md
tauri-apps/tauriReview and test all LLM output before submitting; do not use AI to answer review comments..github/CONTRIBUTING.md
Textualize/richAI-generated PRs accepted if labelled with the agent and linked to a maintainer-approved solution.AI_POLICY.md
Textualize/textualAI-generated PRs accepted if labelled with the agent and linked to a maintainer-approved solution.AI_POLICY.md
vitejs/viteAI welcome under two rules: write comments, issues and PR descriptions yourself, and understand what you submit.CONTRIBUTING.md
vitest-dev/vitestAI may generate code you fully understand; comments, issues and PR descriptions must be in your voice.AI_POLICY.md
wagtail/wagtailPurely AI-generated work is unwelcome; disclose AI use and write the PR description yourself..github/CONTRIBUTING.md

Disclose31 repos

AI-assisted work is accepted if disclosed, and you stay responsible for it.

Repos whose AI policy: AI-assisted work is accepted if disclosed, and you stay responsible for it.
RepoWhat it saysPolicy
antiwork/gumroadEvery PR ends with an AI disclosure naming the model and prompts; current frontier models are recommended.CONTRIBUTING.md
apache/arrowLinked guidance accepts AI PRs if you say what was generated and review it; disengaged PRs get closed.CONTRIBUTING.md
apache/kafkaAdd a Co-Authored-By or Generated-by commit trailer naming the AI tool; the human author stays responsible.CONTRIBUTING.md
caddyserver/caddyAI-assisted code is fine, but every account must disclose LLM use; no automated interactions with people..github/CONTRIBUTING.md
coollabsio/coolifyAI is allowed; mention any AI use in the PR and understand and verify every change.CONTRIBUTING.md
crewAIInc/crewAIAI-authored PRs and issues must carry the llm-generated label or may be closed without review..github/CONTRIBUTING.md
curl/curlAI-assisted code accepted at normal standards; reports from AI tools must say so and be verified first.docs/CONTRIBUTE.md
denoland/denoAI tools carry no penalty, but their use must be disclosed in the PR description..github/CONTRIBUTING.md
expo/expoName the AI tool in the PR, edit agent drafts yourself, and never let agents open PRs alone.CONTRIBUTING.md
getnao/naoAI-written code is fine if understood; state the model used and write the PR blurb yourself.CONTRIBUTING.md
ghostty-org/ghosttyAll AI use must be disclosed and fully understood; AI-generated media is banned; slop authors get denounced.AI_POLICY.md
github/spec-kitAI is encouraged, but disclose tool, model, settings and extent; agents label every commit and comment.CONTRIBUTING.md
go-gitea/giteaAI help is welcome with clear disclosure; review and test it, and don't rely on AI for replies.CONTRIBUTING.md
JuliaLang/juliaDisclose substantive generative AI contributions with details, review every change, and open PRs yourself.CONTRIBUTING.md
llvm/llvm-projectAny tool with a human in the loop; label substantial AI content; no unapproved agents or good-first-issue fixes.llvm/docs/AIToolPolicy.md
microsoft/TypeScriptFully agent-written patches are fine if understood and disclosed; bulk autonomous agent PRs are never accepted.CONTRIBUTING.md
moby/mobyDisclose tools and extent in the description, add a Created with note, and review AI content first.CONTRIBUTING.md
nodejs/nodeAI allowed with disclosure and verification; no automated PRs, AI good-first-issue fixes or wholly AI-written messages.doc/contributing/ai-guidelines.md
ocaml/ocamlWrite or personally review every part; disclose significant AI-generated code, descriptions or comments and the tool.AI.md
openssl/opensslNon-trivial AI portions need an Assisted-by trailer naming agent and model, plus a CLA with AI clauses.CONTRIBUTING.md
oxc-project/oxcDisclose AI use and review and test everything; repeated low-quality AI PRs lead to bans.CONTRIBUTING.md
pnpm/pnpmAgent-made PRs welcome if understood and tested; agent-written PRs, issues and comments need a disclosure footer.CONTRIBUTING.md
PostHog/posthogAI is fine if you understand and test the work; disclose agent involvement in the PR template.AI_POLICY.md
pytorch/pytorchAI-generated content in PRs, issues and comments must be labelled and explained; no fully autonomous agents.AI_POLICY.md
ray-project/rayLinked guidance requires PRs to state AI assistance and test results; agent-only PRs are not allowed.CONTRIBUTING.md
ReactiveX/RxJavaDisclose any AI use in the PR, even for discussion only; you answer for whatever the LLM produced.CONTRIBUTING.md
run-llama/llama_indexHighlight where AI wrote code and how you verified it; prefer AI for small, well-tested changes.CONTRIBUTING.md
sqlfluff/sqlfluffAI-assisted PRs welcome at the same bar; disclose material AI use and what you validated manually.CONTRIBUTING.md
starship/starshipEvery AI-assisted PR must disclose it; no substantial AI on good first issues; no autonomous agents.AI_POLICY.md
storybookjs/storybookAI as a personal assistant is welcome; a real person opens each PR and discloses the tool.CONTRIBUTING.md
TanStack/routerAI drafts are fine if reviewed and owned; agents are told to disclose authorship; no mass PRs.CONTRIBUTING.md

Allows34 repos

AI-assisted work is accepted under the usual quality and responsibility rules, with no disclosure asked.

Repos whose AI policy: AI-assisted work is accepted under the usual quality and responsibility rules, with no disclosure asked.
RepoWhat it saysPolicy
apache/pulsarA human must review, submit and own AI-assisted changes; fully autonomous agent PRs are not acceptable.CONTRIBUTING.md
appwrite/appwriteAI does not lower the bar: you own the PR and review agent output against AGENTS.md.CONTRIBUTING.md
AvaloniaUI/AvaloniaAI-assisted PRs accepted when a human fully understands them; autonomous agents are not; keep descriptions short.CONTRIBUTING.md
dart-lang/sdkReview all AI code, make sure it builds and passes tests, understand it, and verify AI-written text.CONTRIBUTING.md
detekt/detektAI-assisted PRs are fine; work that looks entirely AI-generated with little human review is treated differently..github/CONTRIBUTING.md
directus/directusAI-assisted code allowed if fully understood; no autonomous agents; no wholly AI-written issues, PRs or comments.ai_policy.md
earendil-works/piAI-written code is fine if you understand it; issues should be in your own voice or labelled.CONTRIBUTING.md
erlang/otpIssues must be created by humans; AI may help draft an issue you have reviewed yourself.CONTRIBUTING.md
flutter/flutterReview and understand AI code, verify AI-written text, and never let AI replace human discussion.CONTRIBUTING.ai.md
grafana/grafanaAI help is fine if you review, test and understand everything; keep pull requests focused and small.AI_POLICY.md
GreptimeTeam/greptimedbAI-assisted PRs need end-to-end understanding of the design, with uncertain AI-written parts flagged for reviewers.CONTRIBUTING.md
h4ckf0r0day/obscuraAI-assisted contributions are fine; low-quality or unreviewed agent output will be closed.CONTRIBUTING.md
honojs/honoAI may be used, but contributions that waste maintainer time can be closed and accounts blocked.docs/CONTRIBUTING.md
Kludex/starletteYou may use AI tools but stay responsible for every contribution; duplicate PRs lead to bans.AI_POLICY.md
Kludex/uvicornYou may use AI tools but stay responsible for every contribution; duplicate PRs lead to bans.AI_POLICY.md
ml-explore/mlxAI-assisted coding is allowed; you stay accountable and must refine output to reflect your understanding.CONTRIBUTING.md
MoonshotAI/kimi-codeAI-assisted PRs meet the same standard as hand-written ones; understand and explain what you submit.CONTRIBUTING.md
nautechsystems/nautilus_traderAI help allowed with full human review; disclosure not required; no autonomous agent actions or AI co-authors.AI_POLICY.md
neovim/neovimReview and refine AI output before requesting review; no one-shot AI PRs; cut verbose AI text.CONTRIBUTING.md
penpot/penpotResponsible AI use is supported; AI-assisted content must be human-reviewed and verified before submission.CONTRIBUTING.md
prettier/prettierSubmit only code you understand and have tested, follow the templates; low-quality AI content is closed.CONTRIBUTING.md
prisma/ormNo AI disclosure asked; results are verified, and an agent skill describes the contribution workflow.CONTRIBUTING.md
prometheus/prometheusAI bot PRs need a human owner or two maintainer approvals; every change gets two human reviewers.CONTRIBUTING.md
psf/requestsLLM tools allowed if you own every line; no LLM co-author tags, unsupervised agents or LLM review comments..github/AI_POLICY.md
pydantic/pydanticAI use welcome if you fully understand the code; spammy or incoherent AI PRs may be closed.docs/contributing.md
quarkusio/quarkusAI may assist if you understand and own the results; unsupervised bots and copied AI output are unacceptable.AI_POLICY.md
rollup/rollupAI-created PRs are fine, but maintainers may then edit them freely without prior discussion.CONTRIBUTING.md
serverpod/serverpodAI-assisted PRs are fine if you understand and validate them and answer reviewer questions personally.CONTRIBUTING.md
spring-projects/spring-frameworkAI agent contributions accepted with careful human review; accounts run by autonomous bots are banned.CONTRIBUTING.md
streamlit/streamlitResponsible AI use is welcome with full author ownership; outside pull requests are currently paused.CONTRIBUTING.md
TanStack/aiAI may generate code you understand, test and verify; mass low-quality AI PRs are treated as spam.CONTRIBUTING.md
timescale/timescaledbAI coding assistants are welcome; AI code meets the same standards and low-quality AI PRs are closed.CONTRIBUTING.md
vitessio/vitessAI-written code is fine if it reflects real work and understanding; AI-driven issue sniping is discouraged.CONTRIBUTING.md
zed-industries/zedLLM coding welcome with a human who understands it; no autonomous agents; don't let LLMs write your messages.CONTRIBUTING.md

Welcomes6 repos

The project actively invites contributions made with AI or coding agents.

Repos whose AI policy: The project actively invites contributions made with AI or coding agents.
RepoWhat it saysPolicy
aaif-goose/gooseUse AI however you like with no disclosure needed, but review the code yourself before submitting.CONTRIBUTING.md
ClickHouse/ClickHouseEmbraces AI for development with optional disclosure; review generated code and own it like hand-written code.AI_POLICY.md
cloudflare/vinextBuilt almost entirely by AI; contributors are encouraged to use AI, and every PR gets AI review.CONTRIBUTING.md
lightpanda-io/browserAI-assisted or AI-driven PRs are welcome; summarize in your own words and discuss large changes first.CONTRIBUTING.md
openclaw/openclawAI and vibe-coded PRs are first-class with no disclosure needed; show evidence and understand the code.CONTRIBUTING.md
warpdotdev/warpAny coding agent may implement contributions, with free cloud agents offered; talk to maintainers as a human.CONTRIBUTING.md

Methodology

  • The sample is the 388 GitHub repos on the leaderboard on 16 Nov 2026. GitLab and Bitbucket repos weren't scanned.
  • A keyword search of CONTRIBUTING and AI_POLICY.md-style files found candidates; a few large projects it missed were checked by hand. We read each one in full, along with any policy it links to, and dropped sections that aren't about AI-assisted contributions.
  • Each policy was classified by hand on 16 Nov 2026. Policies change; the linked file is the current word.