CLI Command
beam init
Create or print MCP configuration so coding agents can start Beam locally.
beam init is the bridge between Beam and coding agents. It configures an MCP client to launch beam mcp, which exposes the same core design context, asset, variable, and compare behavior as the CLI.
--print is the safest first command. It prints the manual MCP configuration and does not modify any client files, which is useful for review, copy-paste setup, or unsupported clients.
--client codex targets a supported adapter. Config writes must preserve unrelated settings, validate the final JSON, create a backup before changing files, and show the exact path that changed.
Unsupported clients should never fail silently. Beam should print the manual MCP config and explain where the user can place it in their agent settings.
1beam init --print2beam init --client codex3beam init --json
Run Sequence
- Run
beam init --printfirst and review the generated MCP configuration. - If the client is supported, run
beam init --client codexor the matching client adapter. - Restart the coding agent so it can discover the Beam MCP server.
- Ask the agent to use Beam with a Figma frame URL.
Expected Output
--print should show a JSON block with a beam MCP server whose command is beam and whose args are ["mcp"]. Client-specific setup should report the config path it updated.
Files And Artifacts
- Agent MCP configuration file
- Backup of the client config when Beam writes automatically
Next Step
Run beam mcp manually if you need to verify server startup outside the agent.
Recovery Paths
- If the client is unsupported, place the printed JSON in the client's manual MCP configuration area.
- If the agent cannot find Beam, confirm the global
beamcommand is available in the same shell environment the agent uses. - If automatic config fails, use
beam init --printand apply the configuration manually.
