Problem
Agents and MCP clients using Claude need a way to execute code locally without:
- Separate API key management
- Cloud-only execution (losing repo + env context)
- Additional billing outside existing Claude subscription
This gap limits local-first agent workflows and MCP-based tooling.
Solution: cowork-to-code-bridge — MCP Server for Local Claude Code
cowork-to-code-bridge is a production-ready MCP server exposing Claude Code as a local execution backend.
Key Features
- ✅ No API keys — uses Claude subscription on the user's machine
- ✅ Full local context — agent has access to repos, shell, MCPs, environment
- ✅ MCP standard — JSONRPC 2.0 over stdio (native protocol)
- ✅ Zero network ports — file-based queue, token-authenticated
- ✅ Lightweight — <1MB, minimal dependencies
- ✅ Production-ready — 570 lines + 9 unit tests + comprehensive docs
How It Works
Agent/MCP client config:
{
"providers": {
"claude-code-bridge": {
"type": "mcp",
"command": "cowork-to-code-bridge-mcp",
"args": ["--stdio"],
"env": {"BRIDGE_ROOT": "$HOME/.cowork-to-code-bridge"}
}
}
}
Agent escalates work:
response = agent.escalate(
tool="escalate_to_claude",
request="Debug the API failure and propose fix",
wait_seconds=600
)
Three MCP Tools
- escalate_to_claude — Hand task to Claude Code, get result asynchronously
- run_script — Execute whitelisted scripts directly
- list_bridge_scripts — Discover available scripts
Why This Matters
Official MCP server registry. Being listed = discoverability for all MCP clients.
This fills a gap in the MCP registry: there's no canonical "local code execution" backend. Adding cowork-to-code-bridge as a reference implementation helps MCP clients (agents, IDEs, tools) discover the pattern for local Claude integration.
Reference Implementation
Proposed Action
Add cowork-to-code-bridge to the registry under "Local Execution / Code Backends":
Problem
Agents and MCP clients using Claude need a way to execute code locally without:
This gap limits local-first agent workflows and MCP-based tooling.
Solution: cowork-to-code-bridge — MCP Server for Local Claude Code
cowork-to-code-bridge is a production-ready MCP server exposing Claude Code as a local execution backend.
Key Features
How It Works
Agent/MCP client config:
{ "providers": { "claude-code-bridge": { "type": "mcp", "command": "cowork-to-code-bridge-mcp", "args": ["--stdio"], "env": {"BRIDGE_ROOT": "$HOME/.cowork-to-code-bridge"} } } }Agent escalates work:
Three MCP Tools
Why This Matters
Official MCP server registry. Being listed = discoverability for all MCP clients.
This fills a gap in the MCP registry: there's no canonical "local code execution" backend. Adding cowork-to-code-bridge as a reference implementation helps MCP clients (agents, IDEs, tools) discover the pattern for local Claude integration.
Reference Implementation
Proposed Action
Add cowork-to-code-bridge to the registry under "Local Execution / Code Backends":