Loading ToolBench
Convert any API into an MCP server — REST, SOAP, GraphQL, Database, MCP Bridge. Self-hosted MCP middleware with 257+ pre-built adapters.
AnythingMCP provides 4 demo/informational tools with basic definitions but critical gaps in schema rigor, parameter documentation, and production-grade error handling. All tools are READ_ONLY and informational in nature. Tool descriptions are present but generic; only 1 of 4 tools has a documented input schema; parameter descriptions are entirely missing from the one tool with schema; output schemas are not documented anywhere. The tools appear to be hardcoded demo endpoints rather than dynamically generated connectors, which limits their production value. No evidence of error handling patterns, recovery guidance, or response structure documentation. This server scores as FAIR/POOR, acceptable for a demo/reference implementation but not production-ready for agentic use.
Setup instructions to connect an AI client (Claude, ChatGPT, Gemini, Copilot, Cursor) to AnythingMCP.
How to install and run your own AnythingMCP instance with Docker.
Overview of the pre-built connectors and the connector types you can build.
What AnythingMCP is, what this demo endpoint does, and where to learn more.
No input schemas visible for 3 of 4 tools (anythingmcp_overview, anythingmcp_get_started, anythingmcp_list_connectors). Only anythingmcp_connect_client has a documented input schema.
Parameter 'client' in anythingmcp_connect_client has an enum constraint but no description explaining what the parameter controls or when to use which client option.
No output schemas documented for any tool. LLMs cannot plan downstream operations or extract required fields from responses without knowing the response structure.
Tool descriptions are generic and under 100 characters for most tools. Descriptions lack actionable context about WHEN to use each tool and what the LLM should expect to receive.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 63 | 2025-06-18+ | v2 |
No error handling or recovery guidance. Tool definitions contain no information about what errors can occur, whether they are retryable, or what the agent should do if a tool call fails.