Your memory. Your tools.
Connect your agent
Keep useful context across conversations, projects, and agent harnesses. Most clients connect with a single server URL and a browser sign-in. Choose the tool you work in.
OAuth sign-in
Claude Code ↗
Add the server URL and sign in from the terminal.
OAuth sign-in
Codex ↗
Add the server and log in; token fallback included.
MCP reference
Other MCP clients ↗
Standard OAuth settings for any agent, IDE, or application.
Token setup
Rakazo ↗
Use the built-in Serenity memory provider.
OAuth · qualification pending
Claude on the web ↗
Connector settings and current status.
OAuth · qualification pending
ChatGPT ↗
Connector settings and current status.
How OAuth connections work
Add https://serenity.sire.run/mcp to your agent. It opens a browser tab where you sign in to Serenity, choose one project memory, and approve read-only or read-and-write access. Access tokens are short-lived and renewed automatically with rotating refresh tokens. Each connection expires after 30 days, and you can disconnect any agent from Connected agents.
Choose what your agents share
Each brain is a separate memory space. Connect several harnesses to one brain when they should share context. Choose different brains during consent when projects should stay separate. A server name or repository folder alone does not isolate memory.
The free plan includes one brain; additional brains depend on your plan. Connecting another agent to the same brain does not create another memory space.