One memory for Claude, ChatGPT, Codex and Cursor.
Each AI tool keeps its own memory, so Codex doesn't know what you settled in Claude Code, and switching model means starting over. Levirge Brain is one MCP memory that every client reads and writes: Claude, ChatGPT, Codex, Cursor, VS Code, Gemini CLI and more, plus Obsidian for people. The memory sits outside the model, so it stays with you.
Built-in memory stays with the tool that collected it
| Silo | What you see | What removes it in Brain |
|---|---|---|
| Tool | Codex doesn't know what you settled in Claude Code, so you brief it again. | Every MCP client connects to one endpoint and reads the same knowledge base. |
| Model | Switching provider means starting your context over. | Memory sits outside the model; every model reads the same Vault. |
| Teammate | A colleague's agent repeats an investigation yours finished last week. | One shared Vault per project, with per-user tokens. |
| Agent | A new session can't tell what the last one was doing or still owes. | handoff inbox, work_session and todo on the same server. |
One endpoint, every client
https://brain.levirge.com/mcp over Streamable HTTP. Verify any client with kb_overview (read-only).
Claude Code →
Plugin, two commands.
Claude Desktop →
Plugin marketplace, no terminal.
ChatGPT →
Plugin marketplace, sign in.
Codex →
Plugin marketplace, or a token in config.toml.
Cursor →
Plugin, or one block in mcp.json.
VS Code →
MCP: Add Server, used from Copilot Chat in Agent mode.
Gemini CLI →
One gemini mcp add command.
Windsurf →
Token in the raw MCP config.
Cline →
Remote Servers tab, then a token.
Obsidian →
Two-way sync, for people.
Any MCP client →
Streamable HTTP, OAuth or a bearer token.
Antigravity, OpenCode, pi… →
Per-client guides in the docs.
Hand off by role, not by tool
Handoffs sit in the recipient's inbox until it next connects, so agents in different tools never need to be online together. A real run: an evaluation agent's findings went to a second agent, which shipped two merged fixes. Closed-loop handoff →
One memory across AI tools
Can Claude, ChatGPT, Codex and Cursor share the same memory?
Yes, through Levirge Brain. Each client connects to the same MCP endpoint and reads and writes the same knowledge base, so what you settle in one tool is there in the next.
What happens to my AI memory if I switch models or providers?
With Brain, nothing is lost. The memory sits outside the model, in your Vault, so a new model or provider reads the same knowledge the last one wrote.
Which AI tools can connect?
Claude Code, Claude Desktop, ChatGPT, Codex, Cursor, VS Code, Gemini CLI, Windsurf, Cline, Antigravity, OpenCode, pi.dev and oh-my-pi, plus any other MCP client over Streamable HTTP. People can edit the same knowledge in Obsidian with two-way sync.
Can agents in different tools hand work to each other?
Yes. Handoffs live on the same server, addressed by role, and sit in the recipient's inbox until it next connects, so the two agents never need to be online together.
Can I limit what each tool can see?
Yes. Tokens can be limited to chosen Vaults and made read-only. Personal Vaults are private to their owner and fail closed.
What does it cost?
Personal is $20 a month and Personal Pro $60 a month, both for one user. Teams is $35 per seat a month with a minimum of five seats. Every plan includes unlimited agents and unlimited Vaults.
For Claude Code teams in particular: shared memory for Claude Code →
Stop briefing every tool from scratch
Unlimited agents and Vaults on every plan. Personal from $20 a month; Teams $35 per seat.