Andrezi / Continuum / Mi-Memory
This entry is not a single product, but rather a collection of three adjacent directions of open projects: Andrezi focuses on local memory governance for Claude Code, Continuum focuses on shared code graphs and cross-session memory for multi-code agents, and Mi-Memory focuses on a lifecycle memory framework for personal AI.
Who it is for: Developers using Claude Code and willing to maintain memory content and rules continuously. Developers using Claude Code, OpenCode, Codex, or Gemini CLI collaborating with multiple agents in the same codebase. Teams conducting research or developing products for continuous services across phones, cars, homes, wearables, cameras, and tools for personal AI.
Core capabilities
- Andrezi Cross-Session Memory Governance: Memory is delayed through frozen-snapshot rule and deterministically merged into a structured review for injection into the next session.
- Continuum Shared Code Graph and Agent Memory: A long-lived workspace daemon maintains real-time code graphs and persists architectural decisions, agent intentions, and handoff scratchpads in local SQLite.
- Continuum Hybrid Code Search: Fuses BM25 lexical ranking and local semantic embeddings through reciprocal rank fusion.
- Mi-Memory Multi-Source Evidence Memory: MemSense converts multimodal observations into evidence with sources, and MemFuse connects cross-device related observations into coherent events.
- Mi-Memory Governed Evolution: Incorporates corrections, forgetting, and strategy updates into the lifecycle and constrains accepted evolution through gate/rollback records.
Pricing
No verifiable pricing tiers or packages are provided, still pending verification.
Updated 2026-08-24