Frequently asked questions
Use this page for orientation. Each answer links to the guide with the detailed configuration, protocol, or operational behavior.
Understand ERPBridge
What is ERPBridge?
ERPBridge is middleware that connects existing ERP APIs to AI agents through the Model Context Protocol (MCP). It exposes ERP API operations as discoverable MCP tools without modifying the ERP codebase.
How does the integration work?
Register an ERP API with bridgectl, generate and validate MCP tool schemas, and apply those schemas to the server registry. The server exposes the active tools over streamable HTTP or stdio so MCP clients can discover and call them. See the onboarding guide.
Do I have to modify my ERP?
No. ERPBridge sits at the API boundary and uses the REST APIs you already operate. It does not require ERP code changes, API rewrites, or database migrations. See Connectivity for upstream requirements.
What ERP systems can I connect?
Any ERP or service that exposes a REST API can be registered. On-premises systems, cloud suites, and custom services are supported as long as the server can reach the API and the required authentication can be configured. See Authentication.
Which clients work with ERPBridge?
Any MCP-capable client can connect. Use stdio for local integrations, or streamable HTTP for remote and browser-capable clients. Claude, Cursor, Postman, and custom MCP clients are examples. See Transports.
Products and getting started
What is bridgectl?
bridgectl is the ERPBridge developer CLI. It manages API registrations, tool schemas, registry changes, cache, logs, contexts, API tokens, and shell completion. Start with the bridgectl overview.
Does bridgectl include an agent skill?
Yes. ERPBridge ships the bridgectl-ops agent skill for onboarding, maintenance, authentication, observability, troubleshooting, and sanitized bug reports. Read the skill usage guide for its workflow and safety boundaries.
What does the ERPBridge SDK provide?
The @erpbridge/sdk package provides a typed MCP client, exact-name tool proxy, tool registry and direct-invoke wrappers, log and metrics access, health checks, and cache operations. See the SDK overview and API reference.
Where should I start?
Follow the server quickstart to run the server and make a first MCP tool call. Use the onboarding guide when connecting your own ERP, or the SDK quickstart when embedding the client in a TypeScript application.
Can I try ERPBridge without a live ERP?
Yes. The repository includes a FastAPI mock-erp service and Docker Compose setup for local evaluation. The quickstart and Docker deployment guide use that environment.
Security and operations
How is access secured?
Authenticated routes accept the admin credential or a scoped API token. Tokens can be limited to mcp, metrics, or logs, and token values are shown only once at creation. See authentication and the API token guide.
What observability is available?
The server exposes structured logs, recent-log and streaming endpoints, a health endpoint, and Prometheus metrics. The SDK and bridgectl provide typed or command-line access to those surfaces. See REST API and SDK Logs and metrics.
How does caching work?
Caching is opt-in per tool and supports Redis or a bounded in-memory fallback. Tool schemas can define TTL and flushOn invalidation rules. See Exact Match Caching.
Where can I see what is shipped and planned?
Use the Roadmap for the current status of the Server, bridgectl, and SDK. Shipped, in-flight, and future work have separate labels.
Where can I report a problem or request a feature?
Open an issue in the ERPBridge GitHub repository. Include the component, version, reproduction steps, relevant command or endpoint, and sanitized logs. For operational diagnosis, see Skill Usage.