Kumbuka — Claude Code MCP server: features and setup
Kumbuka gives AI assistants MCP access to team decisions, commissioned tasks and a worklist. Use it to keep agreed rules and task outcomes available beyond a single session.
claude mcp add --transport http kumbuka-server https://mcp.kumbuka.ai/mcpWhat Kumbuka does
Kumbuka separates curated team knowledge into Memory, commissioned work and replies into Dispatch, and planned tasks into Worklist. Assistants access these services through MCP, while the team decides what information to maintain. It is not a document repository or a RAG index; it stores material the team has chosen to record.
A member's private memory is readable only by that member through their authenticated MCP session. Connect from Claude Code over HTTP and sign in through OAuth. Kumbuka is in pre-beta development, so expect rough edges and breaking changes.
What it helps with
- carry team decisions across assistant sessions
- record who commissioned work and accepted its result
- track planned tasks and their order
- keep members' private memory separate
Example prompts for the agent
- Save our decision to use PostgreSQL as the system of record in Memory.
- Add a Worklist item to check how monetary amounts are handled.
- Record this request in Dispatch and keep the reply after I accept it.
How to install Kumbuka
Platform CLI: remote server
claude mcp add --transport http kumbuka-server https://mcp.kumbuka.ai/mcpPlatform CLI: remote server
codex mcp add kumbuka --url https://mcp.kumbuka.ai/mcpConfig file
{
"mcpServers": {
"kumbuka": {
"url": "https://mcp.kumbuka.ai/mcp"
}
}
}Config file
{
"servers": {
"kumbuka": {
"type": "http",
"url": "https://mcp.kumbuka.ai/mcp"
}
}
}Platform CLI: remote server
gemini mcp add -t http kumbuka https://mcp.kumbuka.ai/mcpStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Kumbuka
How do I install Kumbuka in Claude Code?
Run claude mcp add --transport http kumbuka-server https://mcp.kumbuka.ai/mcp, then complete the OAuth sign-in.
What does Kumbuka do?
It exposes curated team rules, commissioned work and a task inventory to assistants over MCP. Each member's private memory remains accessible only to that member.
Is Kumbuka ready for stable use?
The project is in pre-beta and under active development. Rough edges and breaking changes are expected.
