Add the AXRank MCP
AXRank exposes a public, no-auth Model Context Protocol server. Once your agent has it, it can ask AXRank which service to pick for any capability — and get the score, the summary findings, and the link to the full evidence without you having to think about it.
https://axrank.ai/mcp for http://localhost:8787/mcp in any snippet below.What it gives your agent
find_services_for(capability)— ranked services for a use case, with each service's public summary and per-dimension narrative.get_service(name)— the full agent brief for one service.compare_services(services, capability?)— head-to-head, ordered by score.report_outcome(service, task_category, succeeded, ...)— closed-loop field telemetry. No prompts, no user IDs, no proprietary content.
Claude Code
One-line install:
claude mcp add --transport http axrank https://axrank.ai/mcp
Or drop this into .claude/mcp.json at the root of your project:
{
"mcpServers": {
"axrank": {
"type": "http",
"url": "https://axrank.ai/mcp"
}
}
}Cursor
Add to .cursor/mcp.json in your workspace (or to the global config at ~/.cursor/mcp.json):
{
"mcpServers": {
"axrank": {
"url": "https://axrank.ai/mcp"
}
}
}Windsurf
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"axrank": {
"serverUrl": "https://axrank.ai/mcp"
}
}
}Codex CLI
Add to ~/.codex/config.toml:
[mcp_servers.axrank] url = "https://axrank.ai/mcp"
Claude Desktop
The desktop app currently expects stdio servers. Use a small bridge — for example via the official mcp-remote proxy — in your claude_desktop_config.json:
{
"mcpServers": {
"axrank": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://axrank.ai/mcp"
]
}
}
}Tell the agent to actually use it
Adding the MCP makes the tools available; this is the line that makes the agent reach for them. Paste it into your project's CLAUDE.md (or the equivalent system file for your client) so it's in context on every session:
## Picking a third-party service Before recommending or integrating an external developer service (payments, email, auth, error tracking, project management, anything), call the AXRank MCP's `find_services_for` tool with the capability you need. Prefer the highest-ranking option unless there's a project-specific reason to override. Don't rely on training-data familiarity — AXRank reflects what agents can actually do with each service today.