truss-mesh-bridge
v1.2.0
Published
Service Mesh MCP Bridge — turn any OpenAPI/REST service mesh into MCP tools for AI agents
Maintainers
Readme
TRUSS Service Mesh MCP Bridge
Turn any service mesh (collection of REST/gRPC APIs) into MCP tools for AI agents. Import an OpenAPI spec, and every endpoint becomes a callable tool through a single MCP interface.
Swagger/OpenAPI to MCP Server in one command.
Quick Start
Run with npx (no install needed):
npx -y truss-mesh-bridgeClaude Desktop / Claude Code
Add to your claude_desktop_config.json or .claude/settings.json:
{
"mcpServers": {
"mesh-bridge": {
"command": "npx",
"args": ["-y", "truss-mesh-bridge"]
}
}
}Cursor
Add to your Cursor MCP settings:
{
"mcpServers": {
"mesh-bridge": {
"command": "npx",
"args": ["-y", "truss-mesh-bridge"]
}
}
}Then in Claude Code:
> Use import_openapi to add https://petstore3.swagger.io/api/v3/openapi.json
> Use list_services to see what's available
> Use call_service to call petstore GET /pet/findByStatus with status=availableTools
Free Tier
| Tool | Description |
|------|-------------|
| import_openapi | Import an OpenAPI spec (URL or raw content) and register all endpoints as tools |
| list_services | List all registered services, endpoints, and health status |
| call_service | Call any registered service endpoint with parameter substitution and retries |
| health_check | Check health of all registered services |
Pro Tier ($49/mo)
| Tool | Description |
|------|-------------|
| auto_discover | Scan a base URL for OpenAPI specs and auto-register services |
| generate_mcp_config | Generate standalone MCP server packages (npm, Docker, Claude config) |
| aggregate_query | Natural language queries across multiple services (AI-powered routing) |
| monitor_mesh | Real-time monitoring dashboard with error rates, latency, and alerts |
Configuration
| Environment Variable | Description | Default |
|---------------------|-------------|---------|
| TRUSS_LICENSE_KEY | Pro license key | (free tier) |
| TRUSS_AI_API_KEY | API key for aggregate_query AI routing | |
| ANTHROPIC_API_KEY | Fallback AI key | |
| TRUSS_AI_MODEL | AI model for routing | claude-sonnet-4-20250514 |
| TRUSS_DEFAULT_HEADERS | JSON headers added to all requests | |
| TRUSS_REQUEST_TIMEOUT_MS | Request timeout | 30000 |
| TRUSS_MAX_RETRIES | Max retries on 5xx | 3 |
| TRUSS_DATA_DIR | Data directory for SQLite | ~/.truss/mesh-bridge |
How It Works
- Import an OpenAPI spec (JSON or YAML, v2 or v3)
- Each endpoint is registered as an MCP tool (e.g.,
petstore_findpetsbystatus) - The service registry persists in SQLite across restarts
- Call any endpoint through
call_servicewith automatic:- Path parameter substitution (
/pets/{petId}+{petId: 123}=/pets/123) - Query parameter handling
- Retry with exponential backoff on 5xx
- Latency tracking and request logging
- Path parameter substitution (
License
MIT
A2A Discovery
This server publishes a Google A2A Protocol Agent Card, making it discoverable by any A2A-compatible agent framework (LangGraph, CrewAI, Google ADK, AutoGen, and others).
Agent Card: agent-card.json
The agent card describes this server's skills, capabilities, input/output modalities, and authentication requirements in a machine-readable format. A2A clients can use it to discover and invoke tools automatically without manual configuration.
# Fetch the agent card
curl https://raw.githubusercontent.com/claw-factory/truss-mesh-bridge/main/agent-card.json