Every agent gets an identity. Every action gets a receipt.
Every receipt gets a signature. Zero trust by default.
from agentoath import TrustAgent
# Create an agent with cryptographic identity
agent = TrustAgent.create("my-agent", capabilities=["search", "summarize"])
# Every action produces a signed receipt
receipt = agent.record_action("search", {"query": "AI safety papers"})
# Anyone can verify — no central authority needed
assert receipt.verify() # Ed25519 signature checkEd25519 keypair per agent. DID-compatible. No centralized registry required.
Every action produces a cryptographically signed, tamper-proof receipt with full audit trail.
Quantified trust based on receipt history. Configurable decay, category weights, and thresholds.
Agent-to-agent trust delegation with scope limits, expiry, and full chain verification.
Peer-to-peer trust network. Share receipts across organizations without centralization.
Native integrations for LangChain, CrewAI, MCP, and Google A2A. Works with any framework.
pip install agentoath
npm install agentoath
pip install langchain-agentoath
Callback handler, trusted chains, trust-filtered retriever
pip install crewai-agentoath
Trusted crews, step/task callbacks, trust-aware agents
pip install agentoath-mcp-server
Model Context Protocol server with 6 AgentOath tools
pip install agentoath-a2a
Agent-to-Agent bridge with trust headers and ASGI middleware
pip install agentoath-registry
Production-ready trust registry server (FastAPI)
pip install agentoath-shadowos
Trust hooks for tool calls, delegation, memory, messaging