agent-receipts/obsigna

View on GitHub ↗

Agent Receipts — cryptographically signed audit trails for AI agent actions. Protocol spec, SDKs (Go, TypeScript, Python), and MCP proxy.

20 ★3 forksGoUpdated 20d ago

What you need to know

Obsigna is the reference toolset implementing Agent Receipts, an open protocol for producing cryptographically signed, tamper-evident records of AI agent actions via MCP.

Install

brew install agent-receipts/tap/obsigna
obsigna daemon start
For Go: go get obsigna.dev/sdk/go
For TypeScript: npm install @obsigna/sdk-ts
For Python: pip install obsigna

Usage

  • Add PostToolUse hook to ~/.claude/settings.json
  • Inspect audit trail with obsigna list/show/verify
  • Use the MCP proxy to wrap any MCP server

Key features

  • Cryptographically signed audit trails
  • MCP stdio proxy signing every tool call
  • Policy hooks and risk scoring
  • Go/TypeScript/Python SDKs
  • Formal verification of the receipt chain

Best for

Teams that need tamper-evident audit trails for AI agent actions

Caveats

  • Requires the daemon to hold the signing key and own the audit chain
  • In-process trust model allows forgery by anyone with code execution
Platforms: Local · macOSClients: Claude Code · Claude Desktop · Codex
Documentation ↗

Reviewed 2026-08-11

Topics

agent-receiptsaiai-agentsauditcryptographyed25519golangmcpmodel-context-protocolobsignapythonreceiptssecuritytypescriptverifiable-credentialsw3c-vc
Stars
20★
Forks
3
Language
Go
License
Apache-2.0
Created
2026-04-02
Last push
2026-08-15