MCP Scoring
https://mcpscoring.com/mcpScore and test any remote MCP server for protocol compliance, tool definitions, metadata and documentation quality, and search the ranked MCP server directory.
Score breakdown
Lowest: Server metadata 100
Needs attention
All checks passed
Every scored check passed on this run.
Connect
Claude Code, Cursor, VS Code
claude mcp add --transport http mcp-scoring https://mcpscoring.com/mcp
README badge
[](/servers/mcpscoring-com)
5 tools · 2 resources · 1 prompts. Open a tool to test it or chat with it.
Auth
Sent with test calls and chat from this tab only. Never stored.
28 passed
Server metadata
100
Identity, discovery, and usage metadata advertised during the MCP handshake.
8 passing or informational checks hidden. Switch to “All checks” to see them.
Server instructions843 chars
MCP Scoring tests remote MCP servers and keeps a public, ranked directory of them. Typical flows: 1. To audit a server, call score_mcp_server with its public URL. It connects live, runs every check (about 5 to 30 seconds) and returns the score, grade, per section scores and the top issues. 2. To find servers, call search_mcp_directory with a keyword. Each result includes a server_id. 3. For the full breakdown of one server, call get_mcp_server_report with that server_id. 4. To fix a server, call get_mcp_fix_prompt and hand the returned prompt to a coding agent working in that server's repository. 5. To explain how scores are calculated, call list_scoring_checks or read the mcpscoring://methodology resource. Scores are 0 to 100 and graded A to F. Only servers that complete an MCP handshake and expose tools appear in the directory.
Discoverability
100
Client listings and public setup information.
5 passing or informational checks hidden. Switch to “All checks” to see them.
Protocol
100
Connection, authentication, and protocol compatibility.
4 passing or informational checks hidden. Switch to “All checks” to see them.
Tools
100
Tool definitions, annotations, and client compatibility.
14 passing or informational checks hidden. Switch to “All checks” to see them.
MCP Apps
Optional
App resources, metadata, and rendering requirements.
1 passing or informational check hidden. Switch to “All checks” to see it.
Tool documentation
100
How clearly each tool’s title, description, input and output schemas, and behavior annotations guide an agent.
How is the overall score calculated?
The overall score is the average of every applicable, non-optional section score (0–100). MCP Apps is optional and never counts.
Metadata uses weighted points; Discoverability, Protocol and Tools score passed ÷ applicable checks; Documentation averages each tool's weighted title (10%), description (30%), inputs (30%), annotations (15%) and output schema (15%).
Grades: A 85–100 · B 70–<85 · C 55–<70 · D 40–<55 · F <40
Check-by-check across the last 2 runs
| metadata | Stable server name | ✓ | ✓ |
| metadata | Human-readable title | ✓ | ✓ |
| metadata | Useful description | ✓ | ✓ |
| metadata | Server version | ✓ | ✓ |
| metadata | Public website | ✓ | ✓ |
| metadata | Scalable or high-resolution icon | ✓ | ✓ |
| metadata | Server instructions | ✓ | ✓ |
| metadata | Declared capabilities | ✓ | ✓ |
| discoverability | mcp.json or server-card.json | ✓ | ✓ |
| discoverability | Setup instructions at MCP URL | ✓ | ✓ |
| discoverability | ChatGPT | ? | ? |
| discoverability | Claude | ? | ? |
| discoverability | Cursor | ? | ? |
| protocol | MCP negotiation succeeds | ✓ | ✓ |
| protocol | Tools list returns results | ✓ | ✓ |
| protocol | Resources list responds | ✓ | ✓ |
| protocol | Server declares tools capability | ✓ | ✓ |
| tools | All tools have descriptions | ✓ | ✓ |
| tools | Claude tools declare applicable behavior hints | ✓ | ✓ |
| tools | Claude Code metadata fits its 2 KB fields | ✓ | ✓ |
| tools | All tools have OpenAI behavior hints | ✓ | ✓ |
| tools | Claude tool names are within 64 characters | ✓ | ✓ |
| tools | Tool names use valid characters | ✓ | ✓ |
| tools | All tools have a display title | ✓ | ✓ |
| tools | Tool text has no hidden control characters | ✓ | ✓ |
| tools | Cursor tool-name budget advisory | ✓ | ✓ |
| tools | Tool names follow the MCP 1–128 character range | ✓ | ✓ |
| tools | Tool names are unique | ✓ | ✓ |
| tools | Tool input schemas have a valid object structure | ✓ | ✓ |
| tools | Declared output schemas have a valid basic structure | ✓ | ✓ |
| apps | View resources | – | – |
| tools | No prompt-injection directives in tool text | ✓ |