Waymaker
WaymakerDocs
Developer Documentation
Documentation

Using the CLI from an AI Coding Agent

If your AI assistant already has terminal access (Claude Code, Cursor's agent mode, etc.), the simplest way to give it Waymaker access is the CLI itself — no MCP server needed. If instead you want your assistant to call Waymaker as structured tools rather than shell commands, see Connecting an MCP client instead. You can use both; they're not exclusive.

Setup

  1. Install the CLI where the agent runs: npm install -g @waymakeros/cli

  2. Put an API key where the agent's shell can read it — a .env file in the project (add it to .gitignore) is usually simplest:

    # .env
WAYMAKER_API_KEY=wm_sk_your_key_here

  3. Tell the agent it has waymaker available and point it at the command reference.

Partners: per-client key isolation

If you manage multiple client organizations, don't rely on one global key. Every command lands in whichever organization the active WAYMAKER_API_KEY belongs to. Use a separate .env (and a separate API key) per client project directory, and run waymaker auth status before any write operation to confirm which organization you're in.

Scoping what the agent can do

API keys carry individual permissions per resource (commander:read, commander:write, host:read, and so on). If you want an agent to explore without being able to change anything, mint it a key with only the :read permissions ticked. See Managing API Keys.

Next steps