ButlerBrain MCP Server for Claude

Overview

ButlerBrain provides a hosted Model Context Protocol server that adds persistent memory to AI assistants. Claude can store facts, preferences, and project context in ButlerBrain and recall them in later sessions, reducing repetitive re-briefing.

Learn more at butlerbrain.ai. The MCP endpoint is https://api.butlerbrain.ai/mcp.

Features

• Save and retrieve persistent memories across Claude sessions.
• Organise context by projects, tags, or namespaces where supported.
• Hosted remote HTTP transport at api.butlerbrain.ai/mcp.
• OAuth or account sign-in through ButlerBrain connectors.
• Designed for long-running assistant workflows and personal knowledge bases.

Use Cases

• Remember coding conventions and repo preferences for a team project.
• Store user role and stack choices for a multi-week build engagement.
• Recall meeting decisions and action items between weekly check-ins.
• Maintain a rolling glossary of domain terms for a client engagement.
• Persist bug triage context across interrupted debugging sessions.

Installation

Claude (Connectors)

1. Add a custom connector under Settings → Connectors.
2. Enter the MCP URL: https://api.butlerbrain.ai/mcp.
3. Sign in with your ButlerBrain account.
4. Enable the connector and instruct Claude when to read or write memory.

Claude Code

claude mcp add --transport http butlerbrain https://api.butlerbrain.ai/mcp
claude mcp list

Other MCP clients

Add a remote HTTP connector at https://api.butlerbrain.ai/mcp and authenticate with ButlerBrain.

Configuration

{
  "mcpServers": {
    "butlerbrain": {
      "url": "https://api.butlerbrain.ai/mcp",
      "type": "http"
    }
  }
}

Authentication is handled through ButlerBrain sign-in in the connector UI.

Examples

Save project context

Remember in ButlerBrain that this repo uses pnpm, Vitest, and Conventional Commits for all future sessions on Project Atlas.

Recall preferences

What do you remember in ButlerBrain about my preferred TypeScript style for this client?

Update stale memory

Update ButlerBrain: we migrated from Jest to Vitest last week; forget the old Jest setup notes.

Security

• Never store credentials, tokens, or regulated personal data in persistent memory.
• Audit shared workspace memories before inviting new collaborators.
• Delete obsolete memories to reduce leakage if accounts are compromised.

Troubleshooting

Memory not recalled

Confirm the connector is enabled in the current conversation and you are signed into the same ButlerBrain workspace.

Conflicting memories

Issue explicit update or delete instructions; ButlerBrain may retain older entries until overwritten.

OAuth errors

Re-authorise the connector and verify your ButlerBrain subscription includes MCP memory features.

Over-broad retention

Periodically prune memories and avoid storing entire documents when summaries suffice.