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.
By leveraging LangChain, the application isolates runtime dependencies while maintaining low-latency execution traces across multi-step LLM operations.
High-throughput parsing of complex documents (PDFs, Word docs, spreadsheets, presentations)
Structured table, key-value, and metadata extraction directly into agent memory buffers
Automated report generation and document synthesis from multi-source repositories
{
"mcpServers": {
"headroom": {
"command": "python",
"args": [
"-m",
"headroom"
]
}
}
}
Paste into your client configuration file (e.g. ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%/Claude/claude_desktop_config.json on Windows).
Clone headroomlabs-ai/headroom from GitHub
Install dependencies with pip install -r requirements.txt (or uv sync)
Copy .env.example to .env and add your API keys
Run the skill server and connect your AI client via stdio or HTTP
Keeps heavy parsing and raw file structures external to the LLM. Only sanitized excerpts and structured returns are injected into agent memory turns.
Enforce explicit directory allowlists and permission scopes in client configuration to ensure autonomous tool calls never breach system boundaries.
Executes in an isolated local child process communicating strictly over JSON-RPC via stdio with explicit process timeouts.
Headroom is an open-source document capability designed for AI clients and autonomous agents. 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.
Follow the step-by-step installation: 1) Verify your runtime (Python 3.10+), 2) Add the configuration snippet into your AI client settings (e.g. claude_desktop_config.json), 3) Restart the client to initialize tool execution.
Headroom is compatible with all AI tools supporting tool-calling and the Model Context Protocol (MCP), including Claude Desktop, Cursor, Claude Code CLI, Windsurf, OpenCode, and custom agent runtimes.
Headroom executes locally with strict directory allowlists and permission scopes. Heavy parsing and raw file structures stay external to the LLM context window to prevent memory saturation and unauthorized access.
Yes, Headroom is published under the Apache-2.0 open-source license. You can inspect the source code and contribute on GitHub at https://github.com/headroomlabs-ai/headroom.
Explore open-source implementation code, license, and community releases.
Anthropic's official Claude Skills for working with document formats — fill, merge, and extract data from PDFs, Word files, PowerPoint decks, and Excel spreadsheets inside agent workflows.
Reverse engineer anything with agents, from app behavior down to native binaries.
A privacy-focused legal analysis team powered by local Ollama models, Agno, and Qdrant vector search to analyze contracts and legal documents offline.
RAG Tutorials & PipelinesA document question-answering tool powered by local or cloud LLMs with RAG to extract answers, summarize sections, and query uploaded PDF files.
Generative UI & FrontendsDrop files — documents or source code — into a chatbot. The agent extracts entities, concepts, and relationships (or modules, classes, functions, and dependencies), then renders an interactive knowledge graph you can explore.