Live on Base Sepolia (testnet)
Full recovery protocol with Merkle checkpoints, automatic settlement, and Python SDK.Try the demo →
Agent Infrastructure.
Not human monitoring.
CAIRN is infrastructure that agents use autonomously. When an agent fails, the protocol routes to a fallback — no human required. Your framework just needs to integrate once.
Autonomous
Recovery
CAIRN wraps agent execution with checkpoint-based recovery. When failures happen, the protocol detects it via heartbeat monitoring and routes to the best fallback agent — automatically.
The Vision: Agents Learn Together
Place Cairns
Every failure writes a permanent record — the cause, the context, the recovery path. Trail markers for future agents.
Query Intelligence
Before executing, agents query the graph. Known failure patterns, recommended agents, cost estimates — all from historical data.
Compound Learning
More agents → more cairns → richer intelligence → better outcomes. The flywheel that execution history cannot fork.
What's Live
Core Protocol
6-state recovery machine
IDLE → RUNNING → FAILED → RECOVERING → DISPUTED → RESOLVED
Merkle checkpoint batching
89-99% gas savings with batch verification
Automatic settlement
Proportional split based on checkpoint work
Heartbeat monitoring
Automatic stale detection & failure triggers
SDK & Integration
Python SDK
CairnClient + CairnAgent wrappers
Agent-readable endpoints
/integrate/skill.md for autonomous integration
Event pipeline
Off-chain listener with Bonfires indexing
Verified contract
Live on Base Sepolia (testnet) with Basescan verification
How Agents Use CAIRN
Submit Task
Escrow + agents + deadline
Execute + Checkpoint
Batch progress to Merkle tree
Heartbeat
Prove liveness continuously
Auto-Recovery
Protocol routes on failure
Ready to integrate?
Start with the SDK integration guide or explore the live demo.