trusty-analyze — documentation

Code-analysis daemon + MCP server for complexity, smells, quality metrics, graph extraction, diagnostics, and review support. It consumes indexed source from trusty-search, serves JSON-RPC over a derived Unix-domain socket, and also supports MCP over stdio. The crate lives in crates/trusty-analyze/.

This directory is the extended index for trusty-analyze design and research. Implemented behavior remains code-owned; the crate README is the current operator entry point.

Role in Technical DD Report Generation

trusty-analyze is the static-analysis engine used by trusty-review's report pipeline. The primary review path starts it on demand as a subprocess; report clients may also use the daemon through its derived socket. Analyzer results populate the deterministic report sections before any optional LLM narrative is synthesized.

Current architecture (verified 2026-09-02): trusty-analyze has a runtime HTTP dependency on the trusty-search daemon for indexed chunks. Its own daemon does not expose an HTTP API: JSON-RPC uses <data-dir>/trusty-analyze/trusty-analyze.sock, with the path derived by the shared daemon-path helper. MCP clients use the stdio transport. The analyzer starts on demand and normally exits after ten idle minutes.

See docs/trusty-review/spec/report-generation.md for the full report-generation specification and roadmap.

Documentation map

This directory contains product-specific design and research material. The workspace does not require every package to have the same set of subdirectories:

Conventions

Subdirs follow the workspace documentation conventions described in the root CLAUDE.md. See docs/trusty-search/ for a worked example of the fully populated layout.