PromptBranch MCP Server

@promptbranch/mcp is a stdio MCP server exposing the local PromptBranch
library to AI coding agents. Thin adapter over @promptbranch/core; shares
the desktop app's database.
Requires Node.js 22 or later.
Part of PromptBranch · agent workflow: SKILL.md
Install
The MCP server is published as the public @promptbranch/mcp npm package.
Point your MCP client at it:
{
"mcpServers": {
"promptbranch": {
"command": "npx",
"args": ["-y", "@promptbranch/mcp"]
}
}
}
Tools
get_prompt | Fetch a prompt's content (current, or a specific version/branch) |
search_prompts | Full-text search (optional tag/collection filter, result limit) |
list_prompts | List library entries (metadata for browsing) |
report_run | Report an execution outcome: 1–5 rating, summary, metrics |
add_note | Attach a note to a prompt or a specific version |
suggest_variation | Propose an improved version — stays pending until a human approves it in the app |
The MCP server deliberately has no publish or import tool. It can read the
library, log runs and notes, and submit pending suggestions; sharing remains a
human action in the desktop app or CLI.
Configuration
Set PROMPTBRANCH_DB=/path/to.db to point at a different library (the
PromptHub-era PROMPTHUB_DB and the pre-rename PROMPTBUILDER_DB are still
honored as deprecated fallbacks).
Prompts are referenced by title or id; see the
main README for the
"agents propose, humans approve" invariant and onboarding.
Build from source
From the repository root, with Node.js 22 and pnpm 11.7.0 available:
pnpm install
pnpm --filter @promptbranch/mcp build
The built server is dist/index.js. The
MCP integration guide has
configuration and agent-workflow guidance.