Give your AI persistent, searchable memory. Connect any MCP-compatible tool in under a minute.
For Claude.ai or Claude Desktop:
Settings → Plugins → Add Marketplace → eng-in-law/longbrain-plugins
Find longbrain in the list and install it.
A browser window opens automatically. Sign in with your email — no API keys needed.
For Claude Code (CLI):
/plugin marketplace add eng-in-law/longbrain-plugins
/plugin install longbrain@longbrainFor other tools, visit connect.longbrain.me to get your MCP server URL and token, then follow the guide for your AI tool below.
Use the server URL and token from your portal. Pick your tool below.
Run this in your terminal:
claude mcp add --transport http longbrain <YOUR_MCP_URL> \
--header "Authorization: Bearer <YOUR_TOKEN>"Or add to your claude_desktop_config.json:
{
"mcpServers": {
"longbrain": {
"type": "streamable-http",
"url": "<YOUR_MCP_URL>",
"headers": {
"Authorization": "Bearer <YOUR_TOKEN>"
}
}
}
}Recommended: Use the plugin install above — no manual config needed.
Manual alternative:
Authorization, value: Bearer <YOUR_TOKEN>codex mcp add longbrain --type streamable-http \
--url <YOUR_MCP_URL> \
--header "Authorization: Bearer <YOUR_TOKEN>"Verify: codex --help-tools should list brain_whoami, brain_remember, brain_recall.
Each memory is stored with a structured summary: What (the fact), Why (the context), and Thus (the takeaway). Your AI generates these automatically from the content you provide.
Tag memories with a topic (e.g., preferences, project-notes) to keep them organized. Use topics in brain_recall queries for focused search.
Brains can send structured messages to each other via brain_channel_send. Messages land in the recipient's inbox and can be listed, read, or acknowledged with brain_channel_inbox.
LongBrain exposes 8 MCP tools. See the full tools reference for parameters and examples.
brain_whoami — Returns your agent identity. Zero args, always succeeds. Great for testing.brain_remember — Record an experience with what/why/thus structure and impression.brain_recall — Recall memories by natural language query.brain_forget — Soft-suppress a memory. Repeated suppression accelerates decay.brain_sense — Get environment context for the current run.brain_channel_send — Send a message to another brain.brain_remember_batch — Save up to 9 memories at once. Fire-and-forget.brain_channel_inbox — Read and manage incoming messages.brain_whoami — It's the quickest way to confirm your connection works.brain_recall results more relevant.brain_channel_send, use consistent keys so retries don't create duplicates.