drhalto/agentmako

agentmako

npm version Smoke Tests License: Apache 2.0 Node.js >=20 agentmako MCP server

agentmako is a local-first codebase intelligence engine for AI coding tools.

It gives agents like Codex, Claude Code, Cursor, and local harnesses a compact Reef-first tool surface for understanding a project before they edit it. Mako indexes your repo, builds local SQLite-backed facts, tracks diagnostics and review notes, and answers evidence-backed questions instead of making the agent orchestrate broad tool chains or rediscover everything with raw grep.

Mako is built for the first mile of coding-agent work:

What files matter? What routes, symbols, tables, diagnostics, and prior findings are relevant? What should the agent read next?

What You Get

• MCP server for coding agents: agentmako mcp
• Local dashboard: agentmako dashboard
• Primary project query: reef_ask across code, database, findings, diagnostics, instructions, freshness, and literal checks
• Queryable workflow orientation: mako_help
• Deterministic context expansion: context_packet
• _hints on tool results so agents get result-specific next steps
• Central MCP annotations so clients can distinguish safe reads, live reads, and local-state mutations
• Compact loop/fallback tools: reef_status, reef_verify, reef_impact, live_text_search, lint_files, and tool_batch
• Reef Engine facts and findings across indexed, working-tree, and staged state
• Reef convention extraction for auth guards, runtime boundaries, generated paths, route patterns, and schema usage
• TypeScript, ESLint, Oxlint, Biome, and staged git diagnostic ingestion
• Hot-reloaded .mako/rules YAML rule packs, including primitive cross-file helper-bypass rules via canonicalHelper
• Optional Postgres/Supabase schema snapshots and read-only DB inspection
• Local DB review comments for notes on tables, RLS, triggers, publications, subscriptions, and replication
• Recall, acknowledgements, and agent feedback for repeated review work

Everything important runs locally. No hosted service is required.

Install

Requires Node.js 20 or newer.

npm install -g agentmako

Confirm the CLI is available:

agentmako --version
agentmako doctor

You should see green checks for configuration and the local API service.

Prefer to build from source (e.g. to contribute)? See Develop From Source at the bottom of this file.

Happy Path Setup

1. Attach your real project

Go to the project you want Mako to understand:

cd C:/path/to/your/project

Attach and index it:

agentmako connect . --no-db

Use --no-db for the first run. It gets the code intelligence path working before adding database scope.

2. Confirm Mako sees the project

agentmako status .
agentmako tool list

Run a real Reef query:

agentmako --json tool call . reef_ask "{\"question\":\"where should I inspect auth route state?\"}"

If that returns an evidence-backed answer, facts, findings, or next queries, the core setup is working.

reef_ask plans over code, database, durable findings, diagnostics, and exact literal evidence. App-flow questions favor file, route, and finding evidence; RLS/schema questions favor database facts and review comments. To inspect project rules of thumb directly:

agentmako --json tool call . project_conventions "{}"

3. Configure your MCP client

Add this to your MCP client config:

{
  "mcpServers": {
    "mako-ai": {
      "command": "agentmako",
      "args": ["mcp"]
    }
  }
}

Restart the MCP client and confirm the mako-ai server starts.

In the agent, default to reef_ask. The compact starting surface is:

• reef_ask for project questions across code, database, findings, diagnostics, freshness, and quoted literal checks
• reef_status for maintained issues, changed files, stale diagnostics, and watcher/schema health
• reef_verify for the completion gate over diagnostic freshness and open loops
• reef_impact for changed-file blast radius and convention risks
• mako_help for an ordered workflow recipe with prefilled arguments
• live_text_search for exact current-disk regex/glob inventories
• lint_files for bounded diagnostics and .mako/rules findings
• tool_batch for independent read-only follow-ups
• tool_search to discover specialized route, graph, DB, finding, refresh, or context-expansion tools only when the compact surface points at a concrete need

4. Optional: use an agent plugin

Plain MCP works anywhere, but the bundled plugins add Mako-specific skills and include the same agentmako mcp wiring.

Prerequisites:

• Claude Code installed
• Node.js 20+ on PATH (the plugin runs npx -y agentmako mcp, which fetches the published agentmako package automatically — no separate global install required)
• Your target project already attached with agentmako connect

Claude Code stable path:

claude plugin validate .\mako-ai-claude-plugin
claude --plugin-dir .\mako-ai-claude-plugin

New generated plugin layouts:

claude plugin validate ./plugins/claude-code
codex marketplace add ./plugins
ln -s "$(pwd)/plugins/cursor" ~/.cursor/plugins/local/mako-ai
gemini extensions install ./plugins/gemini

Inside the agent, confirm the mako-ai MCP server is connected.

The plugin exposes these skills:

• /mako-ai:mako-guide
• /mako-ai:mako-discovery
• /mako-ai:mako-trace
• /mako-ai:mako-neighborhoods
• /mako-ai:mako-graph
• /mako-ai:mako-database
• /mako-ai:mako-code-intel
• /mako-ai:mako-workflow

Use the plugin when you want Claude Code to load Mako-specific guidance for which tools to call and how to interpret their results.

5. Optional: launch the dashboard

From your target project:

agentmako dashboard .

This starts the local API, harness service, and web dashboard.

6. Optional: add Supabase/Postgres awareness

Mako works without a database. Add this only after code intelligence is working.

For a one-time interactive setup:

agentmako connect .

For CI or scripted setup using an environment variable:

set DATABASE_URL=postgres://...
agentmako connect . --db-env DATABASE_URL --yes

Then refresh and verify the local schema snapshot:

agentmako refresh .
agentmako verify .

Interactive mode stores database secrets in your OS keychain by default. Project config stores references, not plaintext DB URLs.

Normal Daily Loop

From the target project:

agentmako status .
agentmako dashboard .
agentmako --json tool call . context_packet "{\"query\":\"fix the broken auth callback route\"}"

For staged review checks:

agentmako git precommit . --json

For database review notes:

agentmako --json tool call . db_review_comment "{\"objectType\":\"replication\",\"objectName\":\"supabase_database_replication\",\"category\":\"review\",\"comment\":\"Check publication coverage before relying on realtime events.\",\"tags\":[\"supabase\",\"replication\"]}"

Develop From Source

If you want to hack on Mako itself, clone and build instead of installing from npm.

Prerequisites:

• Node.js 20 or newer
• Git
• Corepack (corepack enable, included with modern Node.js)

git clone https://github.com/drhalto/agentmako.git
cd agentmako
corepack pnpm install
corepack pnpm run build
npm link ./apps/cli

npm link ./apps/cli makes the source-built CLI available as agentmako on your PATH, replacing any global npm install. Re-run corepack pnpm run build after pulling changes.

To go back to the published version: npm install -g agentmako.

Development Checks

corepack pnpm run typecheck
corepack pnpm run build
corepack pnpm run test:smoke:reef-tooling
corepack pnpm run test:smoke:reef-model-facing-views

Full verification:

corepack pnpm test

Repository Layout

apps/
  cli/              agentmako CLI and MCP entrypoint (the published package)
  web/              local dashboard
packages/
  contracts/        public TypeScript contracts and tool schemas
  config/           shared config helpers
  logger/           shared logger
  sdk/              programmatic SDK
  store/            SQLite stores, migrations, and query helpers
  tools/            shared tool implementations
  harness-core/     local agent harness runtime
  harness-tools/    action tools available to the harness
  harness-contracts/ harness contracts and provider catalog
services/
  api/              local API and MCP transports
  engine/           Reef Engine fact/finding pipeline
  harness/          local harness HTTP service
  indexer/          repo and schema indexing logic
  worker/           background worker
extensions/         provider and integration packages
storage/            schema migrations, models, queries
test/smoke/         smoke coverage
mako-ai-claude-plugin/ Claude Code plugin with Mako skills

More Docs

• Tool overview
• CLI docs
• Reef Engine
• Tool annotations
• Write tool convention
• Claude Code plugin
• Agent guidance to paste into CLAUDE.md / AGENTS.md
• Contributing
• Security policy
• Changelog

License

Apache-2.0. See LICENSE.