agentmako AGENTS.md

Mako MCP Usage

This file is the short agent-facing guide for projects that have Mako registered as mako-ai:

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

In Claude Code, tools usually appear as mcp__mako-ai__<toolName>. Examples below use bare names. For complete CLI/tool details, use agentmako --help, agentmako tool --help, MCP tools/list, or mako_help.

Operating Model

Mako is deterministic project context, not a replacement for reading files, editing carefully, and running tests. Use it to narrow scope: relevant files, symbols, routes, schema objects, findings, freshness, and risks.

Prefer reef_ask before broad grep when the question is about structure, cross-file impact, routing, auth, database usage, diagnostics, or known findings. Prefer live_text_search or shell rg when exact current disk text matters.

Evidence modes:

• Indexed/Reef: fast and structured, but tied to the last index or persisted fact snapshot.
• Live: current filesystem or database. Use for exact lines, edited files, or new/generated files.

Read _hints on every Mako result. They are generated from the actual output and often give the best next step.

Default Workflow

Default to reef_ask for project questions. It combines codebase, database, durable findings, diagnostics, instructions, freshness, and quoted literal checks without making the agent manually orchestrate broad tool chains:

{
  "question": "what matters if I change tenant-scoped dashboard role checks?",
  "mode": "plan",
  "focus": ["app/dashboard/layout.tsx"],
  "changedFiles": ["app/dashboard/layout.tsx"]
}

When you need an ordered workflow instead of an answer, call mako_help. It returns a recipe with suggested args and verification steps:

{
  "task": "audit auth flow for tenant-scoped dashboard role checks",
  "focusFiles": ["app/dashboard/layout.tsx"],
  "changedFiles": ["app/dashboard/layout.tsx"]
}

When reef_ask needs raw ranked context expansion, call context_packet: