Anythingmcp

What HelpCode-ai/anythingmcp MCP server does

HelpCode-ai/anythingmcp MCP server acts as a self-hosted gateway between MCP clients and external systems. It can represent REST APIs, SOAP services described by WSDL, GraphQL endpoints, and SQL or NoSQL databases as MCP tools. Supported database types include PostgreSQL, MySQL, MariaDB, MSSQL, Oracle, MongoDB, and SQLite. It can also connect to other MCP servers and expose their discovered tools through the gateway.

The project includes pre-built adapters for categories such as logistics, ERP, HR, e-commerce, payments, and public data. The provided material describes examples including Deutsche Bahn, weclapp, Etsy, Shopware, DHL, and Sendcloud, alongside a catalog of 188 adapters.

How it works

HelpCode-ai/anythingmcp MCP server uses imported definitions or live discovery to create runtime tools. Inputs can come from OpenAPI or Swagger documents, Postman collections, cURL commands, WSDL files, GraphQL introspection, custom JSON, or a running MCP server. A visual editor maps tool parameters to request paths, queries, bodies, and headers, and allows tool names and descriptions to be adjusted for the client model.

Tools are registered dynamically, so adding or changing a connector does not require restarting the gateway. The MCP endpoint is exposed separately from the web interface; the default local addresses are http://localhost:3000 for the UI and http://localhost:4000/mcp for MCP traffic.

An optional per-workspace knowledge graph records entity names and relationship metadata inferred from connector definitions and tool calls without storing entity values. Optional AI processing can infer links that heuristics miss, while reusable AI skills can capture tool-use patterns. These features are opt-in and are not required for normal gateway operation.

Setup and configuration

The documented quick start requires Docker 24 or newer, Bash, and OpenSSL. Clone the repository, run its setup.sh script, and open the local web interface to register the first account. That account becomes the administrator. The setup flow can configure deployment mode, domain and HTTPS, secrets, MCP authentication, and optional SMTP or Redis settings.

Manual deployment is also supported by copying .env.example to .env and starting the Docker Compose deployment. The project states that Docker Compose is the normal runtime path and provides a Docker image deployment option. Before exposing an instance publicly, create the administrator account and apply suitable firewall or binding restrictions.

Tools and capabilities

HelpCode-ai/anythingmcp MCP server includes controls for OAuth2 with PKCE or client credentials, Bearer tokens, API keys, Basic authentication, WS-Security, client certificates, LOGIN_TOKEN, and OAuth 1.0a. Connector values can use environment-variable interpolation, with those variables hidden from the model.

Additional governance features include per-tool response mapping, audit records containing inputs, outputs, duration, and status, tool-level whitelisting for custom roles, and per-user MCP API keys. Response mapping can remove selected fields before data reaches the model while retaining the upstream response in the audit record.

Limitations and notes

The gateway is source-available and licensed under AGPL-3.0. AI features are optional rather than required. The supplied material does not identify a single mandatory external API key or a fixed set of environment-variable names, because connector authentication and deployment settings vary by configuration.