opencode-mcp

An MCP (Model Context Protocol) server that lets Claude Code drive an OpenCode instance and delegate work to its subagents — so orchestrator models like Opus or Fable can hand off tasks to the other models OpenCode exposes.

Claude Code
      │
      ▼
MCP Server  (this project)
      │
      ▼
OpenCode SDK / CLI
      │
      ▼
Subagents

Status

Functional. All tools are wired to real OpenCode instances via @opencode-ai/sdk (createOpencodeServer / createOpencodeClient), covered by a Vitest suite. The server also installs SIGHUP/SIGINT/SIGTERM/exit handlers so no opencode serve child process outlives the MCP process.

Related MCP server: claude-code

Design

Task delegation is asynchronous: starting a task returns immediately with a task_id instead of blocking until the subagent finishes. This lets Claude Code fire multiple opencode_start_task calls in parallel — each one opens an isolated OpenCode Session — without hitting MCP client timeouts on long-running work. Status and results are fetched separately via polling.

Tools

Prompts

Configuration

MCP_TOOL_TIMEOUT

opencode_wait_for_task accepts a timeout_ms input, but it's clamped to a server-side maximum so a single call can't block the MCP connection indefinitely. That maximum defaults to 300000 ms (5 minutes) and is configurable via MCP_TOOL_TIMEOUT.

MCP_TOOL_TIMEOUT can be set two ways:

• Environment variable — set it in the MCP server config:

  {
    "mcpServers": {
      "opencode": {
        "command": "node",
        "args": ["/path/to/opencode-mcp/build/src/index.js"],
        "env": { "MCP_TOOL_TIMEOUT": "1200000" }
      }
    }
  }

• CLI argument — pass MCP_TOOL_TIMEOUT=<ms> as an extra arg to the server process:

  {
    "mcpServers": {
      "opencode": {
        "command": "node",
        "args": ["/path/to/opencode-mcp/build/src/index.js", "MCP_TOOL_TIMEOUT=1200000"]
      }
    }
  }

If both are present, the environment variable takes precedence over the CLI argument. Invalid or non-numeric values fall back to the 300000 ms default.

Note that raising this value only raises the server-side clamp — the MCP client's own tool-call timeout (e.g. Claude Code's) must also be set to a value greater than or equal to this max, or the client will abort the call before opencode_wait_for_task has a chance to return.

Project structure

src/
├── index.ts                   # MCP server entrypoint (stdio transport, shutdown handlers)
└── modules/
    ├── tools/                 # One file per MCP tool, registered in index.ts
    ├── prompts/               # One file per MCP prompt, registered in index.ts
    └── shared/                # Cross-tool infrastructure
        ├── server-registry.ts   # Tracks running OpenCode servers; killAllServers() on shutdown
        ├── task-registry.ts     # Maps task_id → OpenCode server + session
        ├── opencode-client.ts   # Builds SDK clients from the registries
        ├── config.ts            # MCP_TOOL_TIMEOUT resolution (env var / CLI arg)
        └── mcp-result.ts        # jsonResult / jsonError MCP output helpers

Each module ships with a colocated *.test.ts Vitest suite.

Getting started

pnpm install
pnpm dev     # runs the server through the MCP Inspector (tsx, no build needed)

Other scripts:

pnpm test           # vitest run
pnpm test:coverage  # vitest run --coverage
pnpm lint           # biome check
pnpm lint:write     # biome check --write
pnpm build          # clean tsc build to ./build (also the typecheck)

To use it from an MCP host, build and point the host at the executable entrypoint:

{
  "mcpServers": {
    "opencode": {
      "command": "node",
      "args": ["/path/to/opencode-mcp/build/src/index.js"]
    }
  }
}