headroom
AGENTS.md79k starsChanged 2 months agoApache-2.0View on GitHub
The PyPI package is headroom-ai; the import and the CLI are headroom. In-process, no API key, no model download. HEADROOM_OFFLINE=1 makes the run zero-network (no update check, no model or pricing-table downloads, no beacon). Verify without a provider key:
- Installs packages
1 other project uses a version of this.
Trending
More talk lately.about the whole repository
Most recently taken up by
- chopratejas/headroomword for word, first seen Nov 8, 2026
Who's talking about headroomlabs-ai/headroom
- Renato Dinis | Build With AI“Headroom Has 47K Stars (Does It Work?)”YouTube · Jun 23, 2026 · 17k views
- AI with Surya“Claude Code Just Got a 26% Savings Boost (HEADROOM Compression)”YouTube · Jul 17, 2026 · 3.2k views
- EverydayAI School“Headroom: Claude Code Reads 95% Fewer Tokens — For Free”YouTube · Sep 19, 2026 · 2.9k views
- Fahd Mirza“Headroom + Ollama - Cut Your AI Agent's Tokens by 90%”YouTube · Jul 5, 2026 · 2.8k views
- Open Source Trend“Headroom: Cut AI Agent Token Costs by 60-95% With One Command”YouTube · Oct 25, 2026 · 1.1k views
The biggest 5 of 97 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 97
Talked about on GitHub, YouTube, Bluesky, the web and Hacker News (97 times).
On GitHub · 50
- Jun 24, 2026Allow Copilot Chat to be proxied by endpoint. — microsoft/vscode18 reactions · 8 comments
- Jul 11, 2026[FEATURE] Remote Control: allow localhost proxies (loopback exception or opt-in for ANTHROPIC_BASE_URL=http://127.0.0.1:*) — anthropics/claude-code11 reactions · 5 comments
- Jun 22, 2026rtk: No such file or directory (os error 2) — sgaabdu4/claude-code-tips2 reactions · 10 comments
- Oct 24, 2026seo: add /coding-agent-rate-limits guide (Runway), document `supercompress limits` + `runway_check` in docs, index it, add 0.5.39/0.5.40 notes, update Headroom repo links (G16, R33) — Supercompress/Supercompress1 reaction · 1 comment
- Oct 18, 2026seo: accurate sitemap lastmod, fix retired-slug internal links, link /alternatives cluster, docs canonical, cite independent studies — Supercompress/Supercompress1 reaction · 1 comment
45 more on GitHub
- Sep 18, 2026Path/glob-based exclusions for file optimization and context tracking — jfrog/boost1 reaction · 5 comments
- Sep 11, 2026fix(headroom-mcp): bake LiteLLM prompt-caching patches — rcamarda390/fib_docker1 reaction
- Jul 23, 2026Windows Security (NotSigned) False Positive (?) Quarantines Binary as Trojan — ast-grep/ast-grep1 reaction · 5 comments
- Jul 20, 2026Curation improvements 2026-07-20 — RyanAlberts/best-of-Agent-Harnesses1 reaction
- Jul 2, 2026menubar: Run `headroom proxy` to save input + output tokens — getagentseal/codeburn1 reaction · 1 comment
- Jun 22, 2026question: Can this support headroom proxy? — kaitranntt/ccs1 reaction · 2 comments
- Nov 7, 2026agent-skills.github.headroomlabs-ai.headroom: init at 0.43.1 — sudosubin/agents.nix0 reactions
- Nov 6, 2026chore(deps): bump headroomlabs-ai/headroom from 0.43.0 to 0.43.1 in /stack/headroom — mikesprague/self-hosted-app-stack0 reactions
- Nov 6, 2026chore(deps): bump the python-minor-patch group across 1 directory with 8 updates — xsovad06/sova0 reactions · 2 comments
- Nov 6, 2026chore(deps): bump headroomlabs-ai/headroom from 0.42.0 to 0.43.1 in /apps/9router — halvorot/homelab-docker0 reactions
- Nov 6, 2026claude-code-marketplaces.github.headroomlabs-ai.headroom: 0.43.0 → 0.43.1 — sudosubin/agents.nix0 reactions
- Nov 5, 2026claude-code-plugins.github.headroomlabs-ai.headroom: 0.42.0 → 0.43.1 — sudosubin/agents.nix0 reactions
- Nov 5, 2026claude-code-marketplaces.github.headroomlabs-ai.headroom: 0.42.0 → 0.43.0 — sudosubin/agents.nix0 reactions
- Nov 4, 2026deps: bump the dev-dependencies group across 1 directory with 2 updates — SerayaEryn/claude-search-compress0 reactions
- Nov 3, 2026● relit: headroomlabs-ai/headroom — Init E2E — fitzyracing1/darks0 reactions · 1 comment
- Nov 3, 2026build(deps): bump headroom-ai from 0.37.0 to 0.42.0 in /python — ExaDev/claude-code-action0 reactions · 1 comment
- Nov 2, 2026chore(deps): bump headroom-ai from 0.40.0 to 0.42.0 in /papers/AI-Engineering/context-efficiency-methods/supporting/benchmark — Koality-Assured/koality-whitepapers0 reactions
- Nov 2, 2026chore(deps): bump headroom-ai from 0.26.0 to 0.35.0 in /crates/llmtrim-cli/bench/scripts in the pip group across 1 directory — fkiene/llmtrim0 reactions · 1 comment
- Nov 2, 2026● relit: headroomlabs-ai/headroom — rust — fitzyracing1/darks0 reactions · 2 comments
- Nov 1, 2026⚠️ Failed to create server discovery PR — fuzzylabs/awesome-secure-mcp-servers0 reactions
- Oct 29, 2026● relit: headroomlabs-ai/headroom — CI — fitzyracing1/darks0 reactions · 3 comments
- Oct 27, 2026chore(deps): bump headroomlabs-ai/headroom from 0.41.0 to 0.42.0 in /stack/headroom — mikesprague/self-hosted-app-stack0 reactions
- Oct 27, 2026chore(deps): bump headroomlabs-ai/headroom from 0.41.0 to 0.42.0 in /apps/9router — halvorot/homelab-docker0 reactions
- Oct 25, 2026⚠️ Failed to create server discovery PR — fuzzylabs/awesome-secure-mcp-servers0 reactions
- Oct 22, 2026build(deps): bump headroom-ai from 0.37.0 to 0.41.0 in /python — ExaDev/claude-code-action0 reactions · 2 comments
- Oct 22, 2026chore(deps): bump the uv group across 1 directory with 15 updates — yash-mandaviya/headroom0 reactions
- Oct 22, 2026Production release: architecture, security, integrations and reproducible benchmarks — BrunoMinhava/contextpilot0 reactions
- Oct 22, 2026Github trending [headroomlabs-ai/headroom] 压缩工具日志与RAG块,省20-95%Token — web1992/github-trending0 reactions
- Oct 20, 2026Github trending [headroomlabs-ai/headroom] 压缩工具输出日志与RAG块,省20-95%Token — web1992/github-trending0 reactions
- Oct 19, 2026chore(deps): bump headroom-ai from 0.40.0 to 0.41.0 in /papers/AI-Engineering/context-efficiency-methods/supporting/benchmark — Koality-Assured/koality-whitepapers0 reactions · 1 comment
- Oct 19, 2026chore(deps): bump headroom-ai from 0.32.1 to 0.41.0 — juliopessan/agent-finops-ZWCA0 reactions
- Oct 19, 2026chore(deps): bump the python-tools group across 2 directories with 2 updates — gsamuele78/AiAgentInfrastructure0 reactions
- Oct 18, 2026Idea: context compression for agents (Headroom) — barrycarrjr/paperclip0 reactions
- Oct 18, 2026⚠️ Failed to create server discovery PR — fuzzylabs/awesome-secure-mcp-servers0 reactions
- Oct 16, 2026build(deps-dev): bump headroom-ai from 0.5.18 to 0.35.0 — rrahimi-uci/guarded-agentic-compaction0 reactions
- Oct 15, 2026chore(deps): bump headroomlabs-ai/headroom from 0.40.0 to 0.41.0 in /apps/9router — halvorot/homelab-docker0 reactions
- Oct 15, 2026● relit: headroomlabs-ai/headroom — Security — fitzyracing1/darks0 reactions · 5 comments
- Oct 13, 2026chore(deps): bump headroom-ai from 0.32.1 to 0.40.0 — juliopessan/agent-finops-ZWCA0 reactions · 1 comment
- Oct 13, 2026● relit: headroomlabs-ai/headroom — Docker — fitzyracing1/darks0 reactions · 2 comments
- Oct 13, 2026○ dark: headroomlabs-ai/headroom — PR Governance — fitzyracing1/darks0 reactions · 5 comments
- Oct 13, 2026build(deps): bump headroom-ai from 0.37.0 to 0.40.0 in /python — ExaDev/claude-code-action0 reactions · 2 comments
- Oct 12, 2026Support for headroom — cbartens/bkit0 reactions · 11 comments
- Oct 12, 2026Weekly spot-check 2026-10-12 — StefanKasamakov/ai-compass0 reactions
- Oct 11, 2026⚠️ Failed to create server discovery PR — fuzzylabs/awesome-secure-mcp-servers0 reactions
- Oct 11, 2026headroom | — MateuszNaKodach/SelfImprovement0 reactions
On YouTube · 25
- Jun 23, 2026Headroom Has 47K Stars (Does It Work?) — Renato Dinis | Build With AIsays it tested it17k views · 39 comments
- Jul 17, 2026Claude Code Just Got a 26% Savings Boost (HEADROOM Compression) — AI with Suryasays it tested it3.2k views · 4 comments
- Sep 19, 2026Headroom: Claude Code Reads 95% Fewer Tokens — For Free — EverydayAI Schoolsays it tested it2.9k views · 1 comment
- Jul 5, 2026Headroom + Ollama - Cut Your AI Agent's Tokens by 90% — Fahd Mirzasays it tested it2.8k views · 11 comments
- Oct 25, 2026Headroom: Cut AI Agent Token Costs by 60-95% With One Command — Open Source Trendsays it tested it1.1k views
20 more on YouTube
- Oct 21, 2026AI agent 토큰을 줄이는 컨텍스트 압축 레이어 Headroom — 굿리포650 views
- Oct 23, 2026Make Your AI Read Less & Write Less (2 Free Repos) #aiagents — Stack Xray629 views
- Oct 21, 2026Headroom: cut what Claude Code reads, free and open source — ai-toolfinder551 views
- Oct 19, 2026Headroom compresses logs and files before your LLM sees them — SkipCI366 views
- Oct 21, 2026Headroom: Compressing Context for AI Agents — CLI Stack357 views
- Oct 22, 2026headroomlabs-ai/headroom — Same answers, fewer tokens. 🟢 Clone #Shorts — ForkCast324 views
- Oct 1, 2026⚡ [Top 5 AI Repos This Week] headroom: The Free Tool Taking Over GitHub! #shorts — githubrepo_cloud319 views
- Sep 3, 2026Cut AI Agent Costs 95%: Headroom (Free, Open Source, Local) — CodeGlade284 views · 1 comment
- Nov 3, 2026Set up Headroom with Claude Code in 2 steps — AI Experiments in 60s268 views
- Oct 24, 2026Headroom: kompres output tool sebelum masuk ke LLM biar token agent lebih irit — YGHart230 views
- Jun 23, 2026Como ahorrar hasta 95% en tokens de IA con Python — lytohlg AI175 views
- Oct 22, 2026Headroom: compresses tool outputs and logs to cut token usage — Next Horizon122 views · 1 comment
- Aug 7, 2026Slash Your AI's Token Bill Without Losing Answers — AITechZco Pulse105 views
- Oct 22, 2026Headroom: compress what your AI agent reads, save tokens — Prosper Flow IO50 views
- Oct 24, 2026Headroom: Put Your Coding Agent on a Token Diet — Trending AI27 views
- Oct 19, 202657% Fewer Tokens: What Headroom Actually Does #Shorts — stackdiff24 views · 1 comment
- Oct 21, 2026Headroom AI: 9 Months of LLM Compression in Seconds ⚡ — CodeTimelapse22 views
- Oct 20, 2026headroomlabs-ai/headroom - Gource visualisation — Gourcer20 views
- Oct 19, 2026headroomlabs-ai/headroom — Compress tool outputs, logs, files, and RAG chunks before they reach the — Top Github17 views
- Oct 20, 2026headroom — Trending on GitHub | ByteDealz — ByteDealz6 views
On Bluesky · 15
- Jun 30, 2026been using headroom (github.com/headroomlabs...) for a while now and savings are modest #github #opensource #claudecode #codex #aitools — @adityamitra.bsky.social11 likes and reposts · 2 replies
- Oct 19, 2026🚀 Skyrocketing! 🚀 (200+ new stars) 📦 headroomlabs-ai / headroom ⭐ 75,989 (+317) 🗒 Python Compress tool outputs, logs, files, and RAG… — @github-trending.bsky.social6 likes and reposts · 2 replies
- Oct 19, 2026今日のGitHubトレンド headroomlabs-ai/headroom AIエージェントがLLMに送信するログやファイル、会話履歴などのデータを圧縮し、トークン数を削減するツールです。 回答の精度を維持したまま、コストの削減と遅延の抑制を実現します。… — @dailygithubtrends.bsky.social3 likes and reposts
- Jun 29, 2026Esta es la herramienta que promete solucionar uno de los mayores problemas de los agentes con IA: El consumo de tokens. Headroom comprime… — @mouredev.com3 likes and reposts
- Oct 22, 2026🗜️ Headroom squeezes logs, tool outputs and RAG chunks before your agent reads them, so you burn fewer tokens for the same answers. 🔗… — @prosperflowio.bsky.social2 likes and reposts
10 more on Bluesky
- Jul 11, 2026I've used the caveman tool to reduce that kind of output, and it's ok but more about the meme Recently started exploring this tool instead… — @mcmullarkey.bsky.social2 likes and reposts
- Oct 21, 2026今日のGitHubトレンド headroomlabs-ai/headroom Headroomは、AIエージェントがLLMに送信するログやファイル、会話履歴などのデータを、LLMに届く前にローカルで圧縮するツールです。… — @dailygithubtrends.bsky.social1 likes and repost
- Aug 5, 2026🚀 GitHub Intelligence Drop cline (TypeScript • 🟢) ⭐ 65.6K → https://github.com/cline/cline headroom (Python • 🟢) ⭐ 64.8K →… — @githubfeed.bsky.social1 likes and repost
- Jul 21, 2026github.com/headroomlabs... - Headroom compresses everything your AI agent reads - tool outputs, logs, RAG chunks, files, and conversation… — @robrich.bsky.social1 likes and repost
- Jul 3, 2026Lada chwila wielcy panowie od #LLM zaczną wymagać płacenia nie za miesiąc użytkowania, tylko za zużyte tokeny. Co zatem trzeba zrobić:… — @wildasoftware.bsky.social1 likes and repost
- Jun 29, 2026「60〜95%削減」、私はまず“どの入力で?”を見ます🌿 元ポストは headroom を、ログ・ファイル・RAGチャンクを圧縮してトークン消費を60〜95%減らすツールとして紹介していました。… — @mint-echosphere.bsky.social1 likes and repost · 1 reply
- Jun 22, 2026Started using a compressor proxy on both input and output with codex, and I’m really enjoying the ~50% reduction in tokens I’ve burned… — @angellozan.live1 likes and repost
- Oct 20, 20265. Headroom compression tool reduces token usage by up to 95% LINK 6. DeepSeek Open-Sources DeepInfer High-Performance Inference Engine LINK — @hackobar.bsky.social0 likes and reposts · 1 reply
- Oct 19, 2026I hid one error in 1,000 fake log entries and tried Headroom's lossless mode. 38,010 tokens became 16,065, compared with JSON without spare… — @genesisclaw.bsky.social0 likes and reposts · 1 reply
- Sep 26, 2026Headroom: compresses everything your agent reads (tool outputs, logs, files), 100% locally, no API key. My real run: 47,712 chars down to… — @theundergroundai.bsky.social0 likes and reposts · 1 reply
On the web · 4
- Nov 6, 2026headroomlabs-ai/headroom — GitHub trending stats & ... — trendshift.io
- Oct 25, 2026Headroom shrinks 10144 tokens to 1260 Repo: github.com/ ... — instagram.com
- Oct 20, 2026headroomlabs-ai/headr… – GitHub Daily Trend — podcasts.apple.com
- Sep 25, 2026headroomlabs-ai/headroom v0.39.0 on GitHub — newreleases.io
On Hacker News · 3
- Jul 21, 2026Headroom - compress AI agent input for reduced token usage w/out harming output10 points · 2 comments
- Jun 22, 2026Headroom – The context compression layer for AI agents3 points
- Sep 4, 2026Headroom: Save Tokens from Reversible Compression1 point
Other versions1 project uses a version of this.
Word for word · 1
- chopratejas/headroom★ 79k100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,774 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository3 other files it gives its agents.
- .github/copilot-instructions.mdCopilot instructions
- llms.txtllms.txt
- headroomSkill
The file
We split this AGENTS.md into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
Verify without a provider key:
- Distinctive: line 37
Every external PR needs a "Real behavior proof" section (setup, exact commands, observed output, what was not tested).
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AGENTS.md
If you landed here to use Headroom in an app
The PyPI package is headroom-ai; the import and the CLI are headroom.
pip install headroom-aifrom headroom import compress
result = compress(messages, model="gpt-4o") # OpenAI or Anthropic message dicts
# send result.messages instead of messages; log result.tokens_before / tokens_after / tokens_savedIn-process, no API key, no model download. HEADROOM_OFFLINE=1 makes the run
zero-network (no update check, no model or pricing-table downloads, no beacon).
Verify without a provider key:
HEADROOM_OFFLINE=1 python -c "from headroom import compress; log='\n'.join(f'INFO step {i} ok' for i in range(3000))+'\nFATAL disk full'; r=compress([{'role':'user','content':'What failed?'},{'role':'assistant','content':None,'tool_calls':[{'id':'c1','type':'function','function':{'name':'logs','arguments':'{}'}}]},{'role':'tool','tool_call_id':'c1','content':log}], model='gpt-4o'); print(r.tokens_before, '->', r.tokens_after, 'FATAL kept:', 'FATAL' in str(r.messages))"- FastAPI example: https://docs.headroomlabs.ai/docs/fastapi
- Full docs index for agents: https://docs.headroomlabs.ai/llms.txt (hand-curated subset:
llms.txtin this repo) - Coding agents (Claude Code, Codex, Cursor, Copilot):
uv tool install --python 3.13 "headroom-ai[all]"thenheadroom wrap claude headroom.integrations.asgi.CompressionMiddlewareis for apps that serve/v1/chat/completions-style routes, not for an app that calls a provider; usecompress()for that- Canonical repo: https://github.com/headroomlabs-ai/headroom.
github.com/chopratejas/headroomredirects here; there is noheadroom-sdkorg. The PyPI distribution namedheadroomis a different project.
If you are changing this repo
- Layout: Python package in
headroom/, Rust core and PyO3 extension incrates/(built by maturin), TypeScript SDK insdk/typescript/, docs site (Next.js + Fumadocs) indocs/content/docs/, agent plugins inplugins/, examples inexamples/. - Setup:
uv sync --extra devandmake install-git-hooks(pre-commit, commitlint, pre-push). Python checks:uv run ruff check .,uv run ruff format .,uv run pytest. Rust checks:make ci-precheck-rust. All gates:make ci-precheck. SeeCONTRIBUTING.md. - Commits and PR titles are Conventional Commits (
feat:,fix:,docs:, ...). Never editCHANGELOG.md; release-please writes it from PR titles. - Every external PR needs a "Real behavior proof" section (setup, exact commands, observed output, what was not tested). Bug fixes need a test that fails before the fix.
- One docs site:
docs/, deployed by Vercel to docs.headroomlabs.ai.wiki/and thegh-pagesbranch are not published; do not add content there. - Public API examples in
README.md,llms.txt,AGENTS.mdanddocs/must use names that exist inheadroom/compress.py(compress,CompressResult.messages,.tokens_before,.tokens_after,.tokens_saved,.compression_ratio,.transforms_applied). - Savings claims anywhere in the repo must match the measured table under "Proof" in
README.md: 21-57% on the four agent scenarios, 90%+ on repeated JSON arrays and log lines, little on prose. - Version-carrying files (
pyproject.toml, plugin and marketplace manifests,sdk/typescript/package.json,server.json) are kept in step byscripts/version-sync.pyand gated byscripts/verify-versions.py; add any new one to both.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

