Set up Claude Code

You need Claude Code installed and an Agiler API key. Start with projects:read and restrict the key to the project you want to inspect.

Make the key available

Set AGILER_API_KEY in the environment where you start Claude Code. For a POSIX shell:

export AGILER_API_KEY='ak_REPLACE_WITH_YOUR_KEY'

Replace the placeholder locally. Your shell may retain commands in history; use your usual secret manager or private environment configuration for regular use.

Add the project configuration

Create .mcp.json in the project root, or merge this agiler entry into its existing mcpServers object:

{
  "mcpServers": {
    "agiler": {
      "type": "http",
      "url": "https://mcp.agiler.io/v1",
      "headers": {
        "Authorization": "Bearer ${AGILER_API_KEY}"
      }
    }
  }
}

Keep the variable reference in the file. Each teammate supplies their own key in their environment. Launch Claude Code from that project directory and review its prompt to enable the project MCP server.

Verify the connection

Run these commands to inspect the registered server:

claude mcp list
claude mcp get agiler

Inside Claude Code, use /mcp, then ask:

Use Agiler’s whoami tool to check my account, then list the projects I can access. Do not make changes.

Continue with your first task and project instructions.

If the server does not connect

Check that Claude Code was launched with AGILER_API_KEY set, that the project server is enabled, and that the JSON uses type: "http". Agiler uses the bearer header here, so an OAuth login is unnecessary.

See Claude Code’s MCP documentation for configuration scopes and environment expansion, or Agiler troubleshooting for authentication and tool errors.