Compresses tool outputs, logs, RAG chunks, files, and history locally before they reach the model, as a Python or TypeScript library, a drop-in proxy, an agent wrapper, or an MCP server.
Use it now
Keep it with up to 9 other tools you're setting up, then share them all as one link.
Try this first
Ask your assistant, with Headroom connected:
- Inline Compression
Use Headroom to compress a message list in Python or TypeScript: ‹what you need›
- Agent Wrapping
Use Headroom to launch Claude Code, Codex, Cursor, Aider, and other agents: ‹what you need›
- Reversible Retrieval
Use Headroom to cache originals: ‹what you need›
Connect it over MCP
Connect Headroom as an MCP server, following its setup instructions at https://github.com/headroomlabs-ai/headroom
Its page: github.com/headroomlabs-ai/headroom
Install the CLI
Install it as its own instructions say.
Its page: github.com/headroomlabs-ai/headroom
Call the API
Auth None needed
Its page: github.com/headroomlabs-ai/headroom
Compare
Press Compare to add it to the tray at the bottom of the window, then pick up to 2 others anywhere on the site.
- Access
- APIMCPCLI
- Auth
- None
- Skills
- 5
Skills
Inline Compression
Compresses a message list in Python or TypeScript with compress(), routing JSON, source code, and prose to a matching compressor.
Compression Proxy
Runs headroom proxy as a local drop-in endpoint that compresses traffic to Anthropic, OpenAI, or Bedrock with no application code changes.
Agent Wrapping
Launches Claude Code, Codex, Cursor, Aider, and other agents routed through Headroom with headroom wrap, and undoes it with unwrap.
Reversible Retrieval
Caches originals locally so the model can call headroom_retrieve over MCP when it needs the full text behind a compressed chunk.
Session Learning
Mines failed sessions with headroom learn and writes corrections to CLAUDE.local.md, CLAUDE.md, AGENTS.md, or another agent file.