About Krewbill
Krewbill provides a shared, client-specific memory of notes approved by people, for teams using Claude, ChatGPT, Cursor, or other MCP clients. Assistants can read and search notes, while proposed additions and stale-note flags remain subject to human review.
Key features
- 01Separate approved-note spaces per client
- 02Read and search approved notes
- 03Suggest notes for human approval
- 04Flag potentially outdated notes
- 05Scoped, revocable connection keys
Use cases
- 01Reuse client tone and rules across chats
- 02Look up project facts and decisions
- 03Review proposed or possibly outdated notes
MCP support
Krewbill is a hosted remote HTTP MCP server that exposes tools for listing spaces, reading and searching approved notes, suggesting notes, and flagging stale ones. Connection keys can be scoped to selected spaces and made read-only.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://www.krewbill.com/api/mcpclaude mcp add --transport http ryanv1304-krewbill https://www.krewbill.com/api/mcp{"mcpServers":{"ryanv1304-krewbill":{"type":"http","url":"https://www.krewbill.com/api/mcp"}}}Open in Cursor{"servers":{"ryanv1304-krewbill":{"type":"http","url":"https://www.krewbill.com/api/mcp"}}}Open in VS Code{"mcpServers":{"ryanv1304-krewbill":{"type":"http","url":"https://www.krewbill.com/api/mcp"}}}{"mcpServers":{"ryanv1304-krewbill":{"serverUrl":"https://www.krewbill.com/api/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://www.krewbill.com/api/mcpAuth requiredTools
Details
From the project READMEKrewbill MCP
Krewbill gives each client a shared AI memory: a space of notes that a person has approved. Claude, ChatGPT and Cursor read those notes over a remote MCP server, so your team stops re-explaining a client's tone, rules and facts in every new chat.
Website: https://krewbill.com
This repository only describes the hosted Krewbill MCP service. The service itself is closed-source.
What it does
- One space per client, holding notes a person has approved.
- Assistants can read a space's approved notes.
- Assistants can only suggest new notes. Nothing is saved to the space until a person approves it.
- Each connection key is limited to the spaces you choose, can be read-only, and can be revoked.
- Accounts, workspaces, roles and invitations are available in the web app.
- Works with any MCP client that supports remote (HTTP) servers. Tested with Claude Code; use with other clients may need a different setup.
Tools
| Tool | What it does |
|---|---|
list_spaces | Lists the spaces your key may read. |
get_space_context | Returns a space's approved notes (voice and rules, key facts, decisions). Assistants call it first, before drafting for a client or project. |
search_memory | Searches approved notes by keyword, in one space or in all spaces your key can read. |
suggest_memory | Proposes a new note. It waits in an approval queue until a person approves it, and never changes shared memory by itself. |
flag_stale | Flags an approved note as possibly out of date for a person to review. It never edits or removes the note. |
A key can cover one space, several, or all of them, and can be read-only. Read-only keys cannot suggest notes.
Connect
- Create a space and a connection key at https://krewbill.com (see https://krewbill.com/connect).
- Krewbill shows your personal server address once, when you create the key. Copy it.
- Add it to your assistant. For Claude Code:
claude mcp add --transport http krewbill <your-server-address>
If your tool asks for a header instead of a key in the address, use the plain server address https://www.krewbill.com/api/mcp and send Authorization: Bearer <your-key>.
Create one key per tool, so you can revoke each separately.
Status
Krewbill is free during early access, until December 30, 2026. Connections authenticate with keys.
Feedback
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.