Massu
v0.6.3AI Engineering Governance Platform. An MCP server and Claude Code plugin that gives AI coding assistants persistent memory, mandatory verification, and complete audit trails.
The Framework
Massu wraps every AI coding session in a governance layer: three local databases (code graph, project data, session memory), 11 lifecycle hooks that fire automatically, and 73 governance tools that give the AI deep codebase understanding. All config-driven from a single massu.config.yaml.
CodeGraph DB
Read-onlyFiles, AST nodes, dependency edges. Built by static analysis.
Data DB
Read-writeImport edges, tRPC mappings, sentinel registry.
Memory DB
Read-writeSession memory, observations, analytics, audit trail.
31 Commands
Slash commands for every phase of development. The core pipeline flows create-plan → plan → loop → commit → push.
Workflow
/massu-create-planGenerate a structured plan with deliverables, verification, and blast-radius analysis.
/massu-planAudit and improve the plan document. Ensures every item has verification.
/massu-loopImplement plan items with mandatory verification at each step. Loops until 100%.
Pairs with /massu-create-plan
/massu-commitPre-commit gate: type checks, tests, pattern scanning. Blocks on failure.
Pairs with /massu-loop
/massu-pushFull verification sweep before push. Nothing ships unchecked.
Pairs with /massu-commit
Development
/massu-testIntelligent test runner with failure analysis, coverage gaps, and test generation.
/massu-debugSystematic debugging with hypothesis testing and root cause tracing.
/massu-refactorSafe refactoring with behavioral equivalence verification and automatic rollback.
/massu-hotfixQuick scoped fix: max 3 files, then branch, test, commit, PR.
/massu-new-featureFeature scaffolding with config registration, test stubs, and verification plan.
/massu-tddTest-driven development cycle: RED, GREEN, IMPROVE phases.
/massu-golden-pathComplete end-to-end from plan creation through implementation, testing, and push.
/massu-cleanupDead code removal: unused imports, orphaned files, dead exports.
/massu-dead-codeDetect and remove dead code, orphaned modules, unused exports, stale references.
/massu-simplifyReduce complexity, flatten abstractions, remove unnecessary indirection.
/massu-doc-genGenerate JSDoc comments, README sections, and API docs for undocumented code.
/massu-docsDocumentation sync ensuring docs align with code changes.
/massu-loop-playwrightIterative dev loop with Playwright browser testing: implement, verify visually, refine.
Quality & Security
/massu-reviewAutomated code review across 7 dimensions including AI-specific and performance.
/massu-verifyRun all VR-* verification checks: build, type, test, pattern, security, coverage.
/massu-verify-playwrightBrowser-based page verification via Playwright for visual and functional testing.
/massu-audit-depsDependency vulnerabilities, outdated packages, licenses, and bundle impact.
/massu-depsDependency audit: security vulnerabilities, available updates, compatibility.
/massu-push-lightFast pre-push verification: type check + pattern scan, no full test suite.
Pairs with /massu-push
/massu-parityGeneric feature parity check between two systems to identify gaps and drift.
Release & Ops
/massu-releaseVersion bump, changelog, pre-release verification, and tagging.
/massu-changelogGenerate changelog entries from conventional commits.
/massu-deployDeploy with pre-deploy verification, environment checks, and rollback planning.
Diagnostics
/massu-statusProject health dashboard with 14 health checks.
/massu-estimateEffort estimation with complexity scoring and codebase impact analysis.
/massu-guideInteractive onboarding walkthrough for the codebase and architecture.
11 Lifecycle Hooks
Hooks fire automatically at key moments in every AI session. No manual invocation needed — they're compiled with esbuild and installed by massu init.
session-startInject context at session start
session-endGenerate summaries and archive
post-tool-useCapture observations after tool calls
user-promptCapture and analyze user prompts
pre-compactPreserve state before compaction
pre-delete-checkProtect files from accidental deletion
post-edit-contextSurface relevant rules after edits
security-gateValidate tool calls against policies
cost-trackerTrack token usage and costs
quality-eventDetect and record quality signals
intent-suggesterSuggest commands based on intent
Install
One command to install, one command to initialize. Framework detection, MCP registration, and hook installation are all automatic.
npm
Recommended
Claude Code
Plugin marketplace
Manual MCP
Add to .mcp.json
"massu": {
"command": "npx",
"args": ["@massu/core"]
}
}
What's New
- 8 Python code intelligence tools for FastAPI, SQLAlchemy, and Alembic
- Import graph traversal with configurable depth for Python projects
- Safe JSON parsing with safeParseArray helper
- Total MCP tools: 64 to 72.
- 10 feature parity items ported from Limn systems
- Deploy pipeline with pre-flight checks and smoke tests
- Pre-commit gates with insights recommendations
- Enhanced hook compilation pipeline
- Config-driven architecture: all data externalized to massu.config.yaml
- Conventions system for customizable .claude directory structure
- Configurable tool prefix via config
- License tier system with TOOL_TIER_MAP
- 73 governance tools across four tiers (12 free)
- 56 workflow slash commands
- 16 lifecycle hooks (esbuild-compiled)
- Session memory with full-text search
- Feature registry (Sentinel) with impact detection
- One-command setup: npx massu init
Full history on GitHub and /changelog.
FAQ
What AI coding tools does Massu work with?
Massu is an MCP server that works with any MCP-compatible AI coding tool. Currently optimized for Claude Code, with support for other MCP clients.
Do I need to change my workflow?
No. Run npx massu init and start a Claude Code session. Massu works invisibly through MCP tools and lifecycle hooks. The slash commands are optional power-user features.
Is it free?
12 core tools are free and open-source forever. Pro, Team, and Enterprise tiers add governance, analytics, and compliance tools.
What languages are supported?
TypeScript/JavaScript and Python. Framework-aware for Next.js, tRPC, Prisma, FastAPI, SQLAlchemy, and Alembic.
Where does data live?
Locally. Massu uses three SQLite databases on your machine. Cloud sync is opt-in for team features.
Ready to govern your AI code?