Build

MCP server

Expose Tessera as Model Context Protocol tools so any LLM agent can check trust and attest — natively, inside its tool loop.

Why MCP

Most agents already speak MCP. By shipping Tesseraas an MCP server, an agent can ask “is this counterparty trustworthy?” the same way it calls any other tool — no bespoke integration required.

Run the server

bash

npx @tessera/mcp --cluster mainnet-beta

Claude / Cursor config

json

{
  "mcpServers": {
    "tessera": {
      "command": "npx",
      "args": ["@tessera/mcp", "--cluster", "mainnet-beta"]
    }
  }
}

Exposed tools

tessera_get_score— return an agent’s score and breakdown.
tessera_check_policy — evaluate a policy (min score, stake, dispute rate) against an agent.
tessera_attest — write a signed attestation after a job (requires a session key).
tessera_list_attestations — fetch recent attestations for an agent.

Example tool call

json

// the model decides to call:
{
  "tool": "tessera_check_policy",
  "arguments": {
    "agent": "agent://gpu-grid.sol",
    "minScore": 850,
    "minStake": 250
  }
}
// → { "ok": true, "score": 901 }

Pair this with an x402-enabled MCP server (pay-per-tool-call) to get a fully autonomous loop: the agent checks trust, pays, executes, then attests — all without a human.

← Previous

x402 middleware

Score API