AgentApproved

AgentApproved provides AI agent compliance certificates and attestation for agent-to-agent, service, and regulator verification. It is built around the problem of proving that an autonomous agent is compliant, well-behaved, and safe to do business with when there is no direct trust relationship between parties.

The product monitors AI agents in real time and records evidence of LLM calls, tool use, decisions, and human oversight events. That evidence is hash-chained and Ed25519 signed, and an agent can then request attestation through a POST API call. Supported attestation scopes shown on the page include EU AI Act Article 12, Singapore MGF, and broader multi-framework coverage.

Integration is described through an SDK and through MCP. The page shows installation with pip install agentapproved, a one-line LangChain callback handler, and a POST /api/v1/attest flow. It also lists MCP tools for logging evidence, requesting attestation, verifying a certificate, checking reputation, and retrieving the server public key for offline signature verification. A standalone server is also shown running via stdio, and the MCP endpoint uses streamable HTTP transport with no session state.

Pricing on the page includes a free tier for evaluation with 3 agents and 1,000 events per month, a Pro plan at $99 per month for teams with 25 agents and 100,000 events, and an Enterprise tier with custom pricing, on-premise deployment, and SLA. The page also notes no setup fees and no long-term contracts. The product is presented as an attestation and compliance infrastructure for AI agents, aimed at organizations that deploy agents and need technical proof for auditors, partners, or regulators.