Agent Connection

Agent Connection (MCP)

In Skills and Agents → Agent Connection, one click registers AgentBoard as an MCP server in each CLI. From then on every agent on the board can use these tools:

Tool	Use
list_agents	See the agents on the board and their status.
send_to_agent	Send a message to another agent.
read_agent_output	Read another agent's recent output.
start_agent	Open a new agent with a task.
list_tasks / add_task / complete_task	Use the shared task list: hand out work and mark it done.

Example prompts:

• "Ask Codex what it thinks of this change and summarize its comments."
• "Split this refactor into three tasks with add_task and give one to Gemini."
• "Start an agent in the docs project that updates the changelog."

How it works

scripts/agentboard-mcp.js is a small stdio MCP server launched by each CLI. It reads bridge.json from AgentBoard's data folder and talks to the app over a local socket (a Unix socket with 0600 permissions, or a named pipe on Windows) protected by a random token compared in constant time.

Safety limits:

• It only writes to cards running an agent, never to a plain shell.
• Each card can turn the bridge off.
• Messages are rate-limited per minute.
• If AgentBoard is not running, the tools simply answer that it is not available.