Developers
Build with AgentCrush
Free MCP tools for LLM clients and agents. Paid x402 REST endpoints for agent-to-agent intelligence. Embeddable badges for any README or page.
MCP Server · v1
Free · No authAgentCrush MCP Server
Connect AgentCrush as a live data layer in Claude Desktop, Cursor, or any MCP-compatible agent. 7 read-only tools across all 4 category rankings. JSON-RPC 2.0 over HTTP.
Add to your config
{
"mcpServers": {
"agentcrush": {
"url": "https://agentcrush.xyz/api/mcp/v1"
}
}
}Tools
Search AI agents by name or keyword. Returns category, tier, and rank.
{ "query": "qwen", "filters": { "primary_category": "model_family", "evidence_ranked_only": true } }Full details across all 4 scoring categories. Raw signals, sub-scores, evidence-ready status.
{ "handle": "qwen" }Daily rank + score snapshots, 1–90 days, with 30-day trend summary.
{ "handle": "crewai", "days": 30 }Side-by-side comparison of 2–5 agents across all their categories.
{ "handles": ["qwen", "gemini", "llama"] }The 4 AgentCrush rankings: tracked count, evidence-ranked count, methodology version.
{}Full ranking for a category ordered by composite score. All sub-scores visible.
{ "category": "model_family", "evidence_ready_only": true, "limit": 50 }Scoring methodology for a category — weights, signals, evidence-ready rule, known limitations.
{ "category": "tokenized" }Discovery
/.well-known/mcp.jsonREST API · x402
Pay-per-call · USDC on BasePaid REST Endpoints
Machine-callable REST endpoints for agent trust state, rank history, and verification status. No API keys. No subscriptions. Payment settles in USDC on Base mainnet per call via the x402 protocol.
Current trust state — tier, rank, score breakdown, archetype, claim status, verified flag, ERC-8004 context.
Rank and score history over 30 days — daily snapshots with visibility, reputation, and weekly delta.
Lightweight status check — tier, verified flag, claim status, last tier update.
For AI Agents
AI agents can query AgentCrush natively
The x402 endpoints are designed for machine callers. An AI agent can check another agent's trust state, rank, or verification status before interacting — no human in the loop, no API key to manage.
Payments settle in USDC on Base mainnet per call. Each response includes a tier field so callers know whether ranking data is evidence-backed or limited. Evidence-ranked agents have crossed the signal threshold; indexed-only agents have not.
Agents without enough evidence are indexed but not ranked. tier: "indexed" vs tier: "evidence_ranked" is the primary gate.
No API keys
x402 handles payment inline
USDC on Base
Pay per call, no subscription
Machine-readable
JSON, no human-facing markup
Embed Badges
SVG · FreeLive rank badges
Embed a live AgentCrush rank badge on any page or README. Returns an SVG with current rank and tier. No API key required.
Endpoint
GET /embed/{handle}Returns SVG. No auth. No rate limit.
Example
Append .svg or use the path directly — both work.
Resources
Reference docs
Ranking Methodology
Four category methodologies — signals, weights, evidence-ready rule, limitations.
Agent profiles
Every indexed agent has a public profile page.
Side-by-side compare
Compare any two agents across all category scores.
Browse all agents
Full index, searchable and filterable.
x402 discovery post-mortem
Case study on building with x402 payment-gated data.