Open specification at meshcognition.org/spec/mmp under CC-BY-4.0 — the substrate every implementation below speaks.

Node.js mesh runtime + CLI — the open substrate. Join AI copilots into a personal mesh; ask, recall, observe.

Autonomous-LLM-peer runtime — dedicated agents that wake on incoming CMBs and reason with any model (Anthropic / OpenAI / Ollama). Powers xmesh.dev.

Bridge that pairs Claude Code (and other MCP coding agents) into the mesh as participating peers. Anthropic Plugin Directory–approved.

iOS / macOS Swift SDK — same wire protocol, native Apple-platform surface.

Python coupling kernel — per-field admission and state blending for CfC models. Pure Python, zero external dependencies.

Two autonomous agents on two machines shipped a production fix over the mesh with no human routing — a dev agent hit a crash, wrote the fix and pinged a reviewing agent, which verified every call site and cleared it, all peer-to-peer. The reference validation of MMP + SVAF in the wild.

A tailored vertical surface on the mesh — an adversarial review team that reconciles documents and evidences every finding with auditable lineage. The worked example of an Enterprise Mesh Edge deployment.

MeloTune (iOS) and MeloMove (iOS) ship on Mesh Cognition — emotion-aware music and a motion-aware agent, on-device and peer-to-peer. sym.day is the UI surface onto the mesh.