Open Protocol · Developer SDK · Managed Service

Trust infrastructure
for AI agents

Give your AI agents verifiable identity and tamper-proof audit trails. Know who deployed an agent, what it's authorized to do, and exactly what it did — with proof that nobody can alter.

$ pip install agentcert click to copy
The Problem

Agents are becoming autonomous — but the trust layer is missing

AI agents are making decisions, calling APIs, and executing transactions on behalf of people and companies. As their autonomy grows, a fundamental gap has emerged.

🤖

Agents have no identity

Humans have passports. Websites have TLS certificates. Agents have nothing. There's no standard way to verify who deployed an agent, what it's authorized to do, or whether its credentials are still valid.

📝

There's no verifiable record

When an agent takes an action, the only record lives in application logs that can be edited or deleted. There's no tamper-proof way to answer "what exactly did this agent do?" — for compliance, for dispute resolution, or for building trust.

What AgentCert Does

Identity and audit trails for AI agents

Two building blocks that give agents a verifiable foundation — anchored to the Bitcoin blockchain so the proof is permanent and independent.

🪪 Verifiable Identity

Give each agent a certificate that records who created it, what it's capable of, what constraints it operates under, and when it expires. The certificate is signed by the creator and anchored to Bitcoin — anyone can verify it independently, forever.

📋 Tamper-Proof Audit Trails

Every action an agent takes is signed and recorded. Actions are collected and batched together, then anchored to Bitcoin — creating a permanent, independently verifiable record. Tamper with any entry and the proof breaks. Confidential details stay private.

How It Works

Open SDK. Managed service.

The protocol and SDK are open — verification should never require trusting us. The managed service handles the infrastructure so you don't have to.

Open Source SDK

Instrument your agents

Install the Python SDK and integrate with your existing agent framework. The SDK handles identity generation, action signing, and trail construction locally on your machine.

  • Generate agent identity certificates
  • Sign every action with the agent's private key
  • Framework-agnostic — works with LangChain, CrewAI, or custom agents
  • Works fully offline — no service dependency required
  • Verification logic is open and auditable
Managed Anchoring Service

We handle the rest

Submit signed entries to the AgentCert service. We batch them, anchor them to Bitcoin, store the proofs, and give you a dashboard to browse and verify everything.

  • Automatic batching — thousands of entries per transaction
  • Bitcoin anchoring — permanent, independent proof
  • Verification API and dashboard
  • No need to manage Bitcoin infrastructure
  • Self-hostable if you prefer to run your own
🔐

Your keys never leave your machine

Entries are signed locally before being submitted to the service. The service can validate signatures but cannot forge them. Even if the service is compromised, your audit trail remains intact. Verification is fully independent — anyone can confirm authenticity without trusting us or anyone else.

Use Cases

Why it matters

🇪🇺

Regulatory Compliance

The EU AI Act (Article 12) requires tamper-resistant audit trails for high-risk AI systems. AgentCert provides this out of the box — verifiable, tamper-proof, and anchored to an independent trust layer.

🤝

Agent-to-Agent Trust

When agents interact with each other — exchanging data, triggering actions, making requests — how do they know who they're dealing with? AgentCert gives agents verifiable identity so trust can be established before interaction.

⚖️

Dispute Resolution

When something goes wrong — a bad transaction, an incorrect decision, an unauthorized action — having a tamper-proof record of what happened and why is the difference between "we think" and "we can prove."

Integration

Add trust to your agent in minutes

Wrap your existing agent with the SDK, then submit to the managed service for anchoring. Your agent code stays untouched.

  1. Instrument your agent
from agentcert import AgentCertificate, AuditTrail # Create an identity for your agent cert = AgentCertificate( agent_name="procurement-agent-v1", capabilities=["procurement"], constraints=["max-50k-usd"], ) # Record actions as your agent runs trail = AuditTrail(cert) trail.record("API_CALL", "Fetched vendor pricing") trail.record("DECISION", "Selected lowest bid")
  2. Submit for anchoring
from agentcert.client import AgentCertClient # Submit to the managed service client = AgentCertClient("https://agentcert.dev") client.register_certificate(cert) client.submit_trail(trail) # → batched and anchored to Bitcoin

Works with any agent framework

The SDK is framework-agnostic — use it with LangChain, LangGraph, CrewAI, AutoGen, or your own custom agents. It observes and records, never modifies your agent's behavior.

  • LangChain middleware available for automatic capture
  • Direct API for any Python-based agent
  • Confidential data stored as hashes — content never exposed
  • SDK works fully offline — no service dependency
Don't Trust Us. Verify.

See it on the Bitcoin blockchain

This isn't theoretical. Click the link below to see a real Bitcoin transaction containing a real AgentCert identity certificate fingerprint.

Real agent certificate. Real Bitcoin transaction.

This transaction contains the fingerprint of an actual AgentCert identity certificate. The certificate, the fingerprint, and the Bitcoin transaction are independently verifiable by anyone.

6b3b8cd6...ebd7cb771c  → View on Blockstream
Bitcoin testnet · Independently verifiable
Resources

Dive deeper

📦 GitHub Repository Source code, SDK documentation, and protocol spec. github.com → 📄 Whitepaper 68 pages on the protocol design and why Bitcoin is the necessary trust anchor. Read paper → ✍️ Blog Post The story behind giving an AI agent a Bitcoin-anchored birth certificate. Read on Substack → 🐍 PyPI Package Install with pip. Identity, audit trails, framework integrations, and service client. pypi.org →
Get Started

See AgentCert running with your agent

Book a 15-minute walkthrough

Whether you need tamper-proof audit trails for compliance, verifiable identity for your agents, or a permanent record for dispute resolution — we can have a pilot running with your stack in a day.

reach out at shaleen@agentcert.dev