Why ProvenLog
AI agents call tools, write files, send messages, and make decisions autonomously. When something goes wrong, you need an answer: what did the agent do, and can you prove it?- Hash chain integrity — SHA-256 linked events, tamper-evident by design
- Merkle tree verification — global integrity proof across all agent chains
- Verifiable exports — portable
.plogevidence packages auditors can verify offline - RFC 3161 timestamping — third-party proof that records existed at a given time
- PII/PHI redaction — 17 built-in patterns for emails, SSNs, API keys, credit cards
- Framework-native — drop-in integrations for LangChain, CrewAI, OpenAI, Anthropic, AG2, LangGraph
- Zero-code options — MCP proxy, OTLP receiver, Claude Code hooks
- Embedded or server — local SQLite with no setup, or a server for multi-agent visibility
Get started
Quickstart
Get up and running in under 3 minutes with zero code changes.
Installation
Install the CLI, Python SDK, or TypeScript SDK.
Python SDK
Explicit instrumentation with the Python client.
TypeScript SDK
Instrument your TypeScript agents.
Core concepts
Events
Understand the ProvenLogEvent schema and action types.
Hash Chains
How SHA-256 hash chains provide tamper evidence.
Merkle Trees
Global integrity proofs across all agent chains.
Operating Modes
Embedded (local SQLite) vs Server mode.