proofofwork-agency/reporecall
View on GitHub ↗Local-first, auto-injecting context + memory layer for Claude Code, Codex and MCP. Per-prompt hook injection, explicit freshness/Trust Contract, AST indexing, call graphs, hybrid search, wiki + architecture dashboard. Everything stays on your machine.
37 ★6 forksTypeScriptUpdated 1mo ago
What you need to know
Local-first context and memory layer for coding agents. Auto-injects relevant code, call graph, wiki, and memory into every prompt via Claude Code hooks, with an explicit freshness Trust Contract (indexedCommit, dirty-file count, staleness banners). Deliberately small 6-tool MCP surface; works fully offline.
Install
npm install -g @proofofwork-agency/reporecall reporecall init && reporecall serve
Usage
- •Run reporecall init/serve, then let the hooks push fresh compressed context into every prompt; use the MCP tools for gaps.
Key features
- ✓Automatic per-prompt injection via hooks
- ✓Hybrid retrieval (intent routing + keyword/semantic search + graph expansion + compression)
- ✓Persistent local memory
- ✓Trust Contract freshness signals
- ✓Lens architecture dashboard (single HTML + JSON export)
Best for
Large, high-churn, or unfamiliar codebases where agents need high-signal context
Caveats
- ⚠npm audit reports 3 high advisories (sharp/libvips CVEs deemed unreachable in this usage)
Clients: Claude Code · Codex · Cline · Aider
Documentation ↗Reviewed 2026-08-11
Topics
ai-codingastcall-graphclaude-codecode-intelligencecode-navigationcode-searchcodebase-searchdeveloper-toolsmcpmodel-context-protocolragretrieval-augmented-generationsemantic-searchtree-sitter
- Stars
- 37★
- Forks
- 6
- Language
- TypeScript
- License
- MIT
- Created
- 2026-03-16
- Last push
- 2026-08-04