Headroom
Latest release v0.40.0 on 2026-10-06 · Actively maintained · 15 releases in 90 days
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
| Latest release | v0.40.0 on 2026-10-06 |
|---|---|
| Maintenance | Actively maintained — pushed within the last 14 days, with 3 or more releases in 90 days |
| Stars | 75,028 |
| Releases in 90 days | 15 |
| Last commit | 2026-10-11 |
| Repository | headroomlabs-ai/headroom |
| Licence | Apache-2.0 |
| Open issues | 416 |
Headroom release notes and changelog
v0.40.0 2026-10-06
Fixed
- Anthropic: preserve failed CCR continuations (94bb055)
- Backends/anyllm: map Anthropic toolchoice none to OpenAI none (d722a64)
- Backends/litellm: align streaming usage tokens with the non-streaming path (90a68e0)
- Backends/litellm: keep the upstream 4xx status on sendmessage errors (e0e41cd)
Faster
- Add savings audit output (9a72bc0)
- Compression/code: avoid a UTF-8 copy per code block in byte-to-char mapping (d5e5534)
- Image: OCR and SigLIP each image once, not once per turn (a8561fb)
- Memory: bound traffic-learner persistedids with the dedup window (7e73438)
Changed
- Proxy: read HEADROOMLICENSE and make usage reporting opt-in
- Drop the crewai extra to remove chromadb from the lockfile
- Bump pyjwt 2.13.0 -> 2.15.1 for CVE-2026-102274 (37b9c46)
- Bump source-map-js from 1.2.1 to 1.2.2 in /docs (e06631a)
v0.39.1 2026-09-26
Fixed
- Proxy: stop the 0.39.0 TPM limiter from refusing large-context requests forever (7968122)
v0.39.0 2026-09-25
Fixed
- Anthropic: reject ping-only SSE from buffered CCR (000fefc)
- Anthropic: stop the spurious compaction warning and drop dead handler code (6e71267)
- Backends/litellm: stop dropping image blocks from /v1/messages requests (577336d)
- Binaries: fail closed when a download carries no sha256 pin (26a2c49)
Faster
- Cache: drop eager full-backend TTL sweep on every new-key store (2fed685)
- Cache: LRU-bound the compression-feedback tool-pattern map (2d10b10)
- Memory: reindex FTS5 a page at a time instead of a record at a time (3e36aef)
- Providers: cache Google/Cohere token counters across requests (cc6a07c)
Changed
- Bump dotenv from 17.4.2 to 18.0.1 in /sdk/typescript (56edfbc)
- Bump ruff from 0.16.7 to 0.16.8 in the pip-minor-patch group (b9ded1a)
- Bump the cargo-minor-patch group with 3 updates (fbbe8b8)
- Bump the npm-minor-patch group across 2 directories with 19 updates (54357ff)
Other MCP projects we track
- Codebase Memory · v0.11.0
- Playwright MCP · v0.0.83
- GitHub MCP · v2.0.2
- FastMCP · v4.1.0
- Activepieces · 0.92.2
- MCP Toolbox · v1.14.0
Maintenance status is derived from public repository activity — release cadence and commit recency. It is not a review, an endorsement, or a security assessment. Star counts measure attention, not quality.