Adds cryptographic receipts and persistent memory to AI agents through 24 MCP tools. Every memory observation and agent action gets signed with Ed25519, producing tamper-proof receipts with input/output hashes, timestamps, and provenance chains. Stores everything in local SQLite with full-text search over an entity-observation graph. Your agent calls memory_context to load what it knows, memory_observe when it learns something, memory_recall to search, and track_action to sign operations. Ships with a CLI for exports, a web dashboard for visualization, and bundle support for portable, verifiable memory you can hand to another agent. Runs entirely local with npx, no API keys or servers. Built for scenarios where you need to prove what an agent did and why, not just log it.
claude mcp add --transport stdio webaesbyamin-agent-receipts -- npx -y @agent-receipts/mcp-server