Does Agent Brain send my code or memories to a cloud service?

No. Agent Brain is self-hosted. Core memory and recall operations run on your machine or infrastructure.

Which agents can use it?

Agent Brain works with supported coding tools and local integrations.

API and integrations

Agent Brain is designed to be used by connected coding agents and local automation. Use the integration surface that matches your client:

Surface	Best for
MCP	Agent clients that discover tools at connection time.
REST	Local scripts, dashboards, and service integrations.
Desktop app	Setup, service status, backups, and day-to-day control.

Public contract

The stable concepts are memory lifecycle and retrieval:

remember durable project knowledge
recall relevant context by natural-language query
search and filter stored memories
fetch or archive a memory by ID
share selected memories with connected agents
inspect service and connection status

Request and response bodies are JSON. Clients should discover the current MCP tool schema or use the versioned REST reference shipped with the installation; do not assume private service ports, deployment URLs, or undocumented fields.

Integration guidance

Keep connection URLs and credentials in local configuration or environment variables, never in committed prompts or screenshots.
Scope memories to a project when the knowledge is not globally reusable.
Prefer recall for agent context and search for human browsing.
Archive stale knowledge instead of silently overwriting the historical record.
Treat IDs and response fields as opaque so additive releases remain safe.

For the complete public contract, see the REST API reference, MCP tools reference, and WebSocket events. Those pages describe local interfaces only; they do not document private deployment addresses or credentials.