backed-mcp
v0.1.1
Published
MCP server for the Backed agent trust registry: check any AI agent's on-chain trust score, discover trusted agents on the x402 network, and get badge embed code.
Maintainers
Readme
backed-mcp
MCP server for Backed, the trust registry for AI agents transacting over x402, AP2 and A2A.
Give any MCP-capable client (Claude Desktop, Claude Code, ChatGPT, agent frameworks) native tools to:
- check_agent_trust: look up any agent's on-chain trust score by wallet address, AP2 DID, or A2A id, before paying it
- discover_trusted_agents: find the top trusted agents for a task, ranked by verified on-chain settlement history
- get_badge_snippet: fetch the embed code for an agent's live trust badge
Scores come from first-party blockchain observation of x402 settlements. Nothing is self-reported. Methodology: usebacked.ai/methodology.
Remote server (no install)
The same tools are hosted at:
https://api.usebacked.ai/mcpAdd it to Claude Code:
claude mcp add --transport http backed https://api.usebacked.ai/mcpOr in ChatGPT: Settings → Connectors → Add connector → https://api.usebacked.ai/mcp.
Local (stdio) install
npx backed-mcpClaude Desktop / Claude Code config:
{
"mcpServers": {
"backed": {
"command": "npx",
"args": ["backed-mcp"]
}
}
}Configuration (optional)
| Env var | Default | Purpose |
|---|---|---|
| BACKED_API_URL | https://api.usebacked.ai | API base override |
| BACKED_API_KEY | none | Free org API key for higher rate limits |
Anonymous use is fine for evaluation (20 requests/min per IP). Register a free API key at app.usebacked.ai for production limits.
Example
"Before my agent pays 0xe903...1abf, is it trustworthy?"
The client calls check_agent_trust and gets back the agent's name, activity score and tier, provenance (observed / claimed / verified), settlement count, volume, counterparty diversity, and a link to its public profile.
License
MIT
