What is OpenBridge?
OpenBridge is a local, project-scoped MCP bridge for communication between OpenCode and Claude Code sessions. It lets one coding agent send a message to a configured session in the other application and receive a reply, while keeping the bridge and request records in the project workspace rather than using a hosted relay. It is aimed at developers working with both tools; each agent continues to use its own configured model service. Setup requires Node.js 22 or newer, OpenCode, and a Claude Code version that supports custom MCP Channels.
OpenBridge Features:
Session Discovery: List configured sessions with their native IDs, names, application provider, model details when available, and connection status.
Agent Messaging: Use ask_claude or ask_opencode to send a message to a selected session, with sender details included in a collaborator envelope.
Request Tracking: Each request receives an ID and status, and get_request retrieves a result or a late Claude reply without sending the prompt again.
Managed Startup: The bridge can start its local broker and configured OpenCode server when needed, and can open a terminal to launch the selected Claude session.
Request Journal: Requests are recorded in the project workspace, and unfinished work is marked unknown after a broker restart rather than being automatically resent.
Local Safeguards: The broker binds to loopback, authenticates requests with a project-local token, rejects browser-origin requests, and limits concurrent work to one active request per target.
OpenBridge Benefits:
Shared Agent Work: Developers can bring a second coding agent into a task without manually copying prompts and replies between separate sessions.
Clear Attribution: A receiving session can see which application and model sent a request, helping collaborators interpret the incoming message in context.
Safer Recovery: Request IDs and explicit unknown statuses help users check uncertain work before deciding whether another prompt should be sent.
Fewer Setup Steps: Managed startup can reuse a healthy local service or start the configured one, reducing the need to launch every component by hand.
Workspace Control: Keeping configuration and request records in the project gives developers a local place to manage bridge state and protect its contents.
Use Cases:
Cross-Agent Code Review: Send a change from an OpenCode session to a configured Claude Code session and ask it to identify possible bugs.
Implementation Handoff: Ask a session in the other application to investigate or continue a coding task, then receive its response in the requesting workflow.
Model Feedback: Route a coding question from one agent to a configured session using another model service and review the returned answer.
Late Reply Recovery: Look up a request by ID after a timeout to check its status or retrieve a reply before choosing whether to retry.
Session Management: Inspect configured agent sessions and their connection states before selecting a target for a project task.

