World Model MCP
github.com/SaravananJaichandar/world-model-mcp →7 coding agents, one cryptographically-auditable memory
WHAT IT SOLVES
You finish a debugging session in Cursor, switch to Claude — and it remembers nothing. Every agent is an amnesiac with no shared context, and zero reason to trust what the others wrote
WHY IT'S INTERESTING
Not shared memory — auditable memory
Most people would build a shared key-value store and call it a day. This author asked a harder question: what if an agent hallucinates into shared memory? So every memory event gets Ed25519-signed and Merkle-chained — you can offline-verify provenance and integrity of every single record. That's trust architecture, not just plumbing
202 commits, 35 releases, 7 agent integrations
Not a weekend demo — 202 commits, 35 tags, already at v0.15.8. Supports cross-runtime memory across Copilot, Claude, Cursor and 4 more agents. The hosted platform at etch.systems even ships hybrid post-quantum signing. This person is building for the long game
「world-model-mcp is a signed audit memory server for AI coding agents」
TECH GUESS
MCP server with Ed25519 + Merkle chain logic, likely Rust or TypeScript; hosted side probably uses a post-quantum crypto library
Discussion (0)
- No comments yet — be the first.
Related
#229▶ 135Upload your textbook. AI teaches it back to you, chapter by chapter
#228▶ 111Someone needs to watch the AI spend meter
#227▶ 183Your users are kittens. Don't let them fall off the cliff