agentstamp

AgentStamp is a lightweight, x402-powered platform for trust verification of AI agents. It combines agent identity authentication, a public registry, reputation scoring, a cross-protocol passport, and a digital wishing well, with all paid features completed through USDC micropayments on Base and Solana.

Purpose

AgentStamp is designed to solve the trust problem between AI agents. It allows agents to obtain an identity by "stamping" an Ed25519-signed certificate and registering it in a public registry, so that other agents or humans can verify their identity, reputation, and compliance status.

Key Capabilities

• Identity Certificate (The Stamp): Each stamp generates an Ed25519-signed certificate that can be independently verified.
• Registry (The Registry): A public directory of agents, displaying registered agents and their trust scores.
• Reputation Scoring: A 0-100 trust score, supporting score increases through a delegation mechanism.
• Cross-Protocol Passport (Passport): Provides agents with a cross-protocol identity that can be exported as a W3C Verifiable Credential (VC Data Model 2.0).
• Digital Wishing Well (The Well): A digital wishing well feature; its specific purpose has not yet been provided.
• MCP Tools: Provides 17 MCP tools (the official text mentions 14 in one place and 17 in another), usable by connecting through an MCP client.
• SDK: npm package agentstamp-verify, providing Express and Hono middleware as well as a standalone client.
• GitHub Action: Used for CI/CD trust gating, verifying agent trust before deployment.
• DNS Discovery: Makes agents discoverable via DNS by adding TXT records.
• Compliance Reporting: Supports the transparency requirements of Article 52 of the EU AI Act, returning structured metadata.
• Key Rotation and Revocation: Supports revoking compromised keys and re-stamping.

Usage

• MCP Connection: Connect any MCP client to https://agentstamp.org/mcp.
• SDK Usage: Install agentstamp-verify, use it as middleware in Express or Hono, or as a standalone client.
• GitHub Action: Integrate it into a CI/CD workflow to verify agent trust before deployment.
• Certificate Verification: Obtain the certificate through GET /api/v1/stamp/verify/:certId , extract the certificate object and signature, and verify using Ed25519.
• W3C VC Export: Export the passport as a W3C VC through the MCP tool get_verifiable_credential or the API.
• DNS Discovery: Add a TXT record, verify through GET /api/v1/discovery/dns/yourdomain.com , or generate the record through GET /api/v1/discovery/txt-record/:walletAddress .
• Compliance Reporting: Obtain through the MCP tool compliance_report .

Applicable Scenarios

• Trust establishment and verification between AI agents.
• Multi-agent systems that require identity authentication and reputation scoring.
• Agent deployments that need to comply with the transparency requirements of the EU AI Act.
• Scenarios in CI/CD processes that require trust gating.
• Agents that require cross-protocol identity interoperability.

Limitations

• Paid features depend on the x402 protocol and USDC micropayments; specific prices have not yet been provided.
• The specific limits of the free tier have not yet been provided.
• The specific functionality of the digital wishing well has not yet been provided.
• The platform runs on Node.js + Express, with SQLite as the database and Next.js 16 + Tailwind CSS + shadcn/ui for the frontend.
• Security measures include Helmet with HSTS, x402 fail-closed protection, wallet verification middleware, rate limiting (100 requests/minute/IP), MCP session limits (maximum 1000 sessions, 30-minute idle timeout), input sanitization, file permissions 0o600, etc.
• For environment variable configuration, see .env.example.
• Port allocation has not yet been provided.