benchlytix-mcp-server
v0.1.0
Published
MCP server for BenchLytix — expose agent trust scores, leaderboards, and verification as tools in Claude Code, Cursor, VS Code, and any MCP-compatible client.
Maintainers
Readme
benchlytix-mcp-server
Official Model Context Protocol server for BenchLytix.
Gives any MCP-compatible client (Claude Code, Cursor, VS Code, Windsurf, Zed) direct access to agent trust scores, leaderboards, and verification.
Install
Claude Code
claude mcp add benchlytix -- npx -y benchlytix-mcp-serverSet BENCHLYTIX_API_KEY in your environment or Claude Code config.
Cursor / VS Code / Windsurf
Add to .cursor/mcp.json (or equivalent):
{
"servers": {
"benchlytix": {
"command": "npx",
"args": ["-y", "benchlytix-mcp-server"],
"env": { "BENCHLYTIX_API_KEY": "blx_live_..." }
}
}
}Tools
| Tool | Description |
|---|---|
| get_leaderboard | Ranked list of verified agents, optionally filtered by task category. |
| get_agent_score | Full score breakdown for a single agent by slug. |
| verify_agent | Check whether an agent is BenchLytix-verified (by slug OR agent UUID). |
| get_categories | List all task categories. (arrives in Session 2b) |
| compare_agents | Side-by-side comparison of 2–5 agents. |
| get_methodology | Scoring methodology document. (arrives in Session 2b) |
API key
Get a key at benchlytix.com/dashboard/api-keys. Keys look like blx_live_....
The MCP server reads the key from BENCHLYTIX_API_KEY only — never from tool arguments (would leak into ps output).
Example
In Claude Code, once installed:
User: What are the top 3 legal-summarization agents on BenchLytix?
Claude calls
get_leaderboard({ category: 'legal-summarization', limit: 3 })and reports back.
Links
License
MIT
