Memtrace Public

Memtrace is a shared memory layer for AI coding agents, built as a bi-temporal graph database that tracks code structure and version history. It serves as a single source of truth that an entire fleet of agents can read from, providing a live code map, coordination, and persistence without relying on any LLM calls. Targeted at teams using multiple AI coding agents (e.g., Cursor, Claude Code, Codex, Windsurf), Memtrace aims to eliminate common problems like forgotten context, merge conflicts, and redundant work. It does this by having agents announce intent before editing, checking the code map for overlaps, and automatically resolving safe conflicts. Key differentiators include a custom-built graph database (MemDB) where time is a first-class citizen—every record knows its validity period, so the system can instantly rewind to any past state. The platform is MCP-native, integrates directly into editors via plugins, and boasts a 97.3% accuracy rate on impact analysis, with 13ms query latency. However, as of July 2026, Memtrace is not yet publicly launched. The website only shows a sign-in flow to join a paced access queue (waitlist). No pricing, no changelog, and no public documentation are available. The claims on the homepage (benchmark numbers, fleet coordination) are self-reported and unverifiable without public access. The privacy-conscious design (no LLM calls, local graph) appeals to teams wary of sending code to third-party APIs.