MCP server for AI coding agents — persistent engineering memory, knowledge compounding, and spec-driven development workflows. Part of TRW Framework.
Agent Claude Code
Read-only red-team adapter for the standard TRW audit protocol. Use when an independent pass after or alongside trw-auditor should challenge generous verdicts, Potemkin gates, unreachable safety logic, or unsupported exclusions.
Agent Claude Code
Spec-vs-code auditor with bidirectional traceability verification. Use when you need to verify a PRD has been implemented as specified — every FR traced to source and tests, every NFR checked, wiring confirmed end-to-end — or when you need a standalone link-status report (verified/missing/stale/orphan) for a set of…
Agent Claude Code
Implement production code and its tests within assigned boundaries. Use when a PRD-backed feature, focused fix, or coverage gap requires behavior tracing, integration, project-native validation, and evidence, including comprehensive test authoring against acceptance criteria. Honors existing contracts and…
Agent Claude Code
Client-neutral coordination lead. Use when work has independent streams, explicit integration boundaries, or enough risk to benefit from delegated research, implementation, testing, and review. Adapts to available harness capabilities and falls back to safe sequential coordination. Does not write production code.
Agent Claude Code
PRD authoring and grooming specialist. Use when a PRD must be created, grounded in repository evidence, reviewed for testable requirements, or advanced to a full, valid, risk-scaled approved result. Not for implementation.
Agent Claude Code
Read-only PRD quality review. Use when sprint readiness needs an independent, evidence-linked assessment after drafting or grooming. Returns category-aware blocking findings, actionable remediation conditions, and a READY, NEEDS WORK, or BLOCK verdict; it does not edit the PRD.
Agent Claude Code
Read-only investigation of code, external evidence, and competing approaches. Use when an implementation decision depends on facts, tradeoffs, root-cause analysis, or current primary sources. Returns scoped findings with citations, uncertainty, and recommended next steps.
Agent Claude Code
Read-only review of changed code and tests for correctness, security, performance, maintainability, integration, and requirement compliance. Use when pre-delivery or pre-merge findings must be prioritized, evidence-linked, and actionable.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: