# TeamShared > SuperAgent Memory: shared context for AI swarms and agentic teams. TeamShared is the shared memory layer every agent on the team reads before it answers. Durable semantic recall, reusable skills, named playbooks, shared work, and the graph — one org context, over remote MCP, not locked in a single chat. TeamShared is one remote MCP server: https://teamshared.com/mcp. Pick your agent, paste the prompt, and it adds the server and runs the OAuth sign-in — the same emailed one-time code as the console. No install script, no API key. ## Connect - [MCP server](https://teamshared.com/mcp): Remote MCP. Sign in with OAuth, the same emailed one-time code as the console. No install script and no API key. - [Sign in](https://teamshared.com/login): One-time code to your email. The first sign-in creates a private org. - [Org MCP URL](https://teamshared.com/docs): Use the org's own URL, https://teamshared.com/o//mcp (Copy connector URL in the console org menu), so every session starts in that org. The bare URL uses the org picked at sign-in, or the one marked Make MCP default on All organizations. - [Claude.ai & Desktop](https://teamshared.com/docs/claude-desktop): Open Settings → Connectors → Add custom connector. Name it TeamShared and paste https://teamshared.com/mcp as the URL. Click Connect and finish the emailed one-time code in the browser. In a chat, ask Claude to call the health tool. - [Claude Code](https://teamshared.com/docs/claude): Add the TeamShared MCP server at https://teamshared.com/mcp to Claude Code (run `claude mcp add --transport http teamshared https://teamshared.com/mcp`, user scope). Use OAuth sign-in (no API key). Then run the sign-in flow, tell me when to finish the emailed one-time code in the browser, and confirm by calling the `health` tool. Run /mcp: teamshared should show as connected. Then ask Claude to call the health tool. - [Cursor](https://teamshared.com/docs/cursor): Add the TeamShared MCP server at https://teamshared.com/mcp to Cursor by adding a `teamshared` entry with the `url` field to ~/.cursor/mcp.json (no headers). Use OAuth sign-in (no API key). Then run the sign-in flow, tell me when to finish the emailed one-time code in the browser, and confirm by calling the `health` tool. Open Settings → Tools & MCP: teamshared should be enabled with tools listed. Then ask the agent to call health. - [Codex](https://teamshared.com/docs/codex): Add the TeamShared MCP server at https://teamshared.com/mcp to Codex (run `codex mcp add teamshared --url https://teamshared.com/mcp`), then run `codex mcp login teamshared`. Use OAuth sign-in (no API key). Then run the sign-in flow, tell me when to finish the emailed one-time code in the browser, and confirm by calling the `health` tool. Run codex mcp list (teamshared should be listed), then ask Codex to call the health tool. - [ChatGPT](https://teamshared.com/docs/chatgpt): Open Settings → Connectors → Advanced and turn on Developer mode (menu names shift; look for custom connectors). Choose Create, name it TeamShared, paste https://teamshared.com/mcp, and set authentication to OAuth. Click Connect and finish the emailed one-time code in the browser. In a chat, enable the connector and ask ChatGPT to call the health tool. - [VS Code / Copilot](https://teamshared.com/docs/vscode): Add the TeamShared MCP server at https://teamshared.com/mcp to VS Code by adding a `teamshared` entry (`type: http`, `url` only, no headers) to .vscode/mcp.json, or run the `MCP: Add Server` command. Use OAuth sign-in (no API key). Then run the sign-in flow, tell me when to finish the emailed one-time code in the browser, and confirm by calling the `health` tool. Run MCP: List Servers: teamshared should be running. In Copilot Chat (agent mode) ask it to call the health tool. - [Windsurf / Devin Desktop](https://teamshared.com/docs/devin): Add the TeamShared MCP server at https://teamshared.com/mcp to Windsurf (Devin Desktop) by merging a `teamshared` entry into ~/.codeium/windsurf/mcp_config.json. The remote field must be `serverUrl`, not `url`. Use OAuth sign-in (no API key). Then run the sign-in flow, tell me when to finish the emailed one-time code in the browser, and confirm by calling the `health` tool. Refresh MCP servers in Cascade: teamshared should list tools. Then ask Cascade to call the health tool. - [Gemini CLI](https://teamshared.com/docs/gemini): Add the TeamShared MCP server at https://teamshared.com/mcp to Gemini CLI (run `gemini mcp add --transport http teamshared https://teamshared.com/mcp`), then start `/mcp auth teamshared`. Use OAuth sign-in (no API key). Then run the sign-in flow, tell me when to finish the emailed one-time code in the browser, and confirm by calling the `health` tool. Run /mcp in Gemini CLI: teamshared should be CONNECTED. Then ask it to call the health tool. - [Gemini](https://teamshared.com/docs/gemini-web): On gemini.google.com, add a custom MCP app. Name it TeamShared and paste https://teamshared.com/mcp as the server URL. Start connect. Gemini registers itself. Finish the emailed one-time code. Approve the screen. It names the app and gemini.google.com. Ask Gemini to call the health tool. - [Devin](https://teamshared.com/docs/devin-web): On app.devin.ai, open Customize → MCPs → Connect. Add TeamShared with server URL https://teamshared.com/mcp. Devin registers itself. Finish the emailed one-time code in the browser. Approve the screen. It names the app and app.devin.ai. Ask Devin to call the health tool. - [Grok Bot](https://teamshared.com/docs/grok-bot): In Grok Bot, open the connectors (MCP servers) screen and choose to add a custom server. Paste https://teamshared.com/mcp as the URL and choose Connect. Finish the emailed one-time code in the browser; it hands you back to Grok Bot. Ask Grok Bot to call the health tool. - [OpenClaw](https://teamshared.com/docs/openclaw): Add the TeamShared MCP server at https://teamshared.com/mcp to OpenClaw (`mcp_servers.teamshared`). Use OAuth sign-in if OpenClaw supports it; otherwise ask me for my TeamShared API key (from https://teamshared.com/app/keys) and send it as `Authorization: Bearer `. Set timeout 120 and connect_timeout 30, restart OpenClaw, and confirm by calling the `health` tool. Run openclaw mcp list, then ask the agent to call the health tool. - [Hermes](https://teamshared.com/docs/hermes): Add the TeamShared MCP server at https://teamshared.com/mcp to Hermes (`mcp_servers.teamshared` in ~/.hermes/config.yaml, or `hermes mcp add teamshared --url ... --auth header`). Use OAuth sign-in if Hermes supports it; otherwise ask me for my TeamShared API key (from https://teamshared.com/app/keys) and send it as `Authorization: Bearer `. Set timeout 120 and connect_timeout 30, disable Hermes's built-in memory toolset (`hermes tools disable memory`), reload MCP, and confirm by calling the `health` tool. Run hermes mcp test teamshared, then ask the agent to call the health tool. ## Memory protocol Once connected, each turn: call `memory_session_ensure` (session), then `memory_recall` before non-trivial work (recall), do the work (shared work tools included), and commit the turn with `context_commit` (`close=true` when the task is done). - [Docs](https://teamshared.com/docs): Connect steps for every harness ## Pages - [Docs](https://teamshared.com/docs): Connect an agent - [Connect](https://teamshared.com/connect): The same connect guide as Docs - [Pricing](https://teamshared.com/pricing) - [Blog](https://teamshared.com/blog) - [Privacy](https://teamshared.com/privacy) - [Terms](https://teamshared.com/terms) - [Full connect guide](https://teamshared.com/llms-full.txt): The Connect page as Markdown