Crabeye Mcp Bridge
Collapse many MCP backends into one stdio connection with discovery, namespacing, and execution policies for cleaner agent configs.
Overview
io.github.crabeye-ai/crabeye-mcp-bridge is a Build-phase MCP server that unifies multiple tool backends behind one connection with discovery, namespacing, and execution policies.
What is this MCP server?
- Single MCP connection that fronts multiple underlying tools
- Tool discovery and namespacing to reduce collisions across servers
- Execution policies to gate how tools run from the agent
- npm package @crabeye-ai/crabeye-mcp-bridge v1.3.0 with stdio transport
- npm package @crabeye-ai/crabeye-mcp-bridge version 1.3.0
- Stdio transport per MCP server schema 2025-12-11
Community signal: 9 GitHub stars.
What problem does it solve?
Too many separate MCP entries make agent configs fragile and encourage tool-name clashes without centralized execution rules.
Who is it for?
Power users running several MCP servers who want one consolidated entry and clearer tool governance.
Skip if: Minimal setups with only one or two MCP servers where a bridge adds unnecessary indirection.
What do I get? / Deliverables
You register one bridge server and route namespaced, policy-aware tool access instead of maintaining a growing list of raw MCP connections.
- Single stdio MCP endpoint in your agent config
- Namespaced tool surface for multiple backends
- Policy-governed execution path for agent tool calls
Recommended MCP Servers
Journey fit
Builders hit MCP sprawl while assembling agent stacks, so the primary shelf is Build agent-tooling. Agent-tooling fits a bridge that orchestrates how tools are exposed and executed, not a single app feature or launch task.
How it compares
MCP aggregation layer, not a replacement for domain-specific servers like GitHub or Postgres MCP.
Common Questions / FAQ
Who is crabeye-mcp-bridge for?
Developers and indie builders operating multi-server MCP setups in Claude Code, Cursor, or similar clients who need cleaner discovery and namespacing.
When should I use crabeye-mcp-bridge?
Use it when you are scaling agent-tooling during Build and your MCP config file is hard to maintain or tools conflict by name.
How do I add crabeye-mcp-bridge to my agent?
Install @crabeye-ai/crabeye-mcp-bridge from npm, configure stdio in your MCP settings, wire your downstream tools per Crabeye docs, and point the client at the bridge only.