# Supabrain

> Supabrain is persistent, local-first memory for AI coding agents. It captures the decisions, fixes, and dead ends your agents produce and lets any later agent recall the right one in a single query — across sessions, tools, models, and machines. It is MCP-native and works in Claude Code, Cursor, and any MCP client. Supabrain is in early access.

## Core pages
- [Supabrain home](https://supabrain.io/): What Supabrain is, how it works (capture → recall → compound), the value model, pricing, and FAQ.
- [How to give Claude Code persistent memory](https://supabrain.io/how-to-give-claude-code-persistent-memory.html): Why Claude Code forgets across sessions, why a bigger context window is not memory, and how to add persistent memory over MCP with Supabrain.

## Key facts
- Local-first: embeddings and search run on your machine; nothing leaves unless you turn on sync.
- Model-portable: your memory lives outside the model, so switching models costs nothing.
- MCP-native: one connection over the Model Context Protocol; the agent saves and recalls memories as tool calls.
- Remembers reasoning, not just code — including ruled-out hypotheses (negative knowledge).
- Pricing: Solo is free; Team is a flat 25 USD per person per month with unlimited memories and queries; there is no usage meter.
- Value model (internally audited, grounded in a live 2,000+ memory store across 46 projects, two machines, six months): about 19 days per person per year reclaimed; 31,000–119,000 USD per year at team scale. Time figures are modeled; team figures are projections.
- Re-orienting an agent with no memory typically burns 30,000–150,000 tokens before it does useful work.

## Contact
- Email: hello@supabrain.io