Give your AI true long-term memory. Squish delivers durable context that survives session restarts with semantic search, belief extraction, and automatic memory decay.
Persistent memory gives AI agents the ability to remember past interactions, decisions, and user preferences across sessions. Without it, every session starts from scratch.
Squish derives beliefs (decisions, constraints, preferences) from raw observations automatically. Memory decay archives stale context so relevant memories stay visible.
39 ops/sec throughput, 6.6ms embedding, 6.1ms semantic search. All running locally with SQLite — zero external dependencies required.
Vector databases store raw embeddings. Squish is a memory runtime that adds belief extraction, memory decay, semantic search, and structured recall on top of local storage.
Yes. The self-hosted Squish runtime is free and open-source under MIT. Squish Cloud adds sync and team features starting at $9/mo.
Yes. Squish works across Claude Code, Cursor, Codex, ChatGPT, OpenCode, and any MCP-compatible client from a single runtime.
Yes. All core operations — remember, recall, context — work fully offline with the local SQLite database.