headroom
MCP Serverheadroomlabs-ai/headroom
Headroom is a local context compression layer that reduces LLM token consumption by 60% to 95% for JSON and 15% to 20% for coding agents without losing accuracy. It operates as a library, local proxy, MCP server, and direct agent wrapper to optimize both incoming prompts and outgoing model responses.
66.2k 5.1kApache-2.0Updated 2026-08-13
92
ExcellentOverview
Headroom acts as an intelligent intermediary between your application or AI agent and LLM providers. By processing prompts, tool outputs, logs, and RAG chunks locally through dedicated content routers, structural crushers, AST code compressors, and specialized local models, it significantly cuts down token overhead. It also includes reversible content retrieval, cross-agent shared memory, automated session learning, and output token reduction strategies to minimize overall API expenses while maintaining task accuracy.
Capabilities
- ▸Content-aware token compression for JSON, AST code, and prose
- ▸Local proxy server supporting OpenAI and Anthropic API endpoints
- ▸MCP server implementation for tool integration
- ▸Reversible compression with on-demand original content retrieval
- ▸Cross-agent shared memory and automated learning from session history
- ▸Output token reduction through verbosity steering and effort routing
Best for
Compressing large tool outputs and code search results before sending them to LLMs, Wrapping popular coding agents like Claude Code, Cursor, and Codex to minimize token consumption transparently, Managing cross-agent shared memory and automatically learning prompt corrections from failed sessions, Optimizing outgoing model responses and reasoning effort to reduce high output-token costs
Works with
Claude CodeCodexCursorGitHub CopilotOpenCode
Evoa Score breakdown= Σ (score × weight)
92Task usefulness20%100 → +20.0
実タスクにどれだけ役立つか（機能の豊富さ・用途の明確さ）。 — AIによるcapabilities/use-cases解析
Code quality15%90 → +13.5
実装・指示の品質。 — AIによるSKILL.md/README解析
Maintenance15%100 → +15.0
リポジトリがどれだけ活発に保守されているか。 — GitHub 最終push日時の新しさ
Documentation12%58 → +7.0
ドキュメントの充実度・分かりやすさ。 — README/独自要約の情報量
Security15%100 → +15.0
危険・不審な挙動が無いか。 — AIによるセキュリティレビュー
Originality10%85 → +8.5
ありふれたラッパーではない独自性。 — AIによる独自性判定
Popularity8%100 → +8.0
コミュニティの採用度。 — GitHub Stars/Forks（対数スケール）
Compatibility5%100 → +5.0
対応AIエージェントの広さ。 — AIによる対応エージェント判定
Weighted total92.0 / 100
ライセンス不明/制限あり（Red）のSkillは総合スコアに0.85倍の補正を適用します。 ランキングはこのScoreのみで決まり、広告で変わりません。 算出方法の詳細 →
Security considerations
The application runs locally, ensuring that raw prompts, files, and tool outputs stay on the user's machine before compression. Communication occurs primarily via local proxies and standard API connections.
Categories
Summary and analysis are original content generated by AI Skills Rank. The skill's source text is not reproduced here — view it on the linked repository.