What is PowerMemory?

PowerMemory is a Memory-as-a-Service vault that gives an AI assistant persistent, structured memory about you that survives across sessions and across different AI models. Instead of starting each conversation cold, an assistant connected to PowerMemory loads the relevant approved context you've established — how you want it to communicate (CALIBRATION), what it must never do (HARD_CONSTRAINTS), and your projects, decisions, lessons, preferences and facts — and picks up from where you left off.

Memories are categorized. The assistant reads and writes them through the connector's tools: memory_read / memory_write, plus the session-lifecycle commands — v_power_start opens the session and returns the prior handoff, v_power_save proposes durable new memories for you to approve, v_power_refresh re-anchors mid-session, and v_power_stop seals the session and hands off to the next.

Because the vault is reached over MCP, context saved or approved through one connected assistant can be read by another, so you don't have to repeatedly re-explain established context when you switch tools. This keeps you in control of your own context and lets the assistant work from an accurate picture of you from the first message.

Security & privacy

Vault data is encrypted at rest using SQLCipher with AES-256. PowerMemory is not a zero-knowledge system: its server-side application can decrypt and process vault content as needed to provide the service and return memories to authenticated connected clients. Interoperability with an AI product does not imply endorsement by, or partnership with, that product.

See the FAQ and Privacy + Security Reference for more.