Foxmayn Frappe CLI: command line and MCP server for Frappe

Not an official Frappe project. Not affiliated with Frappe Technologies.

What it does

ffc talks to a Frappe or ERPNext site through its REST API. You can read, create, update, submit and cancel documents from the terminal. It is for administrators, developers and anyone who wants an AI assistant such as Claude to work on their ERP.

It always acts with the permissions of the Frappe user you sign in as. It cannot do more than that user can do in Desk.

Features

• Documents: list-docs, get-doc, create-doc, update-doc, submit-doc, cancel-doc, bulk-update.
• Workflows: workflow apply.
• Search and print: search, pdf.
• Files: files upload, files download.
• Setup: init creates a site profile, whoami shows who you are signed in as.
• Safety flags: --dry-run shows what would change. --read-only blocks every write.
• MCP server: ffc mcp starts a Model Context Protocol server for AI assistants. It is also shipped as an .mcpb bundle and is listed in the MCP registry as io.github.nasroykh/ffc. Use ffc mcp install and ffc mcp uninstall to set up supported clients.
• Desktop app: ffc desktop install and ffc desktop status. The desktop app runs on Windows and macOS. It is a beta pre-release (0.11.x) and is unsigned.
• Updates: ffc update.

Install

Linux and macOS:

curl -fsSL https://raw.githubusercontent.com/nasroykh/foxmayn_frappe_cli/main/install.sh | sh

Windows PowerShell:

irm https://raw.githubusercontent.com/nasroykh/foxmayn_frappe_cli/main/install.ps1 | iex

Homebrew:

brew install nasroykh/tap/ffc

Scoop:

scoop bucket add foxmayn https://github.com/nasroykh/scoop-bucket
scoop install ffc

Then connect your first site:

ffc init
ffc whoami

Releases cover Linux, macOS (amd64 and arm64) and Windows (amd64 and arm64). Checksums are signed with Ed25519, and the install script verifies them when OpenSSL 3 is present. CI runs go vet and go test before each release, and the release workflow publishes a provenance attestation.

Why we built it

We work on Frappe sites every day. Clicking through Desk to check one record or update fifty is slow. A scriptable command line is faster, and it fits into shell scripts and CI jobs.

AI assistants make this more useful, and also more risky. The MCP server gives them real access to a site, but only with one user's permissions, and the --read-only mode lets you start with reading alone. We use ffc in our own AI work on ERPNext.

The tool is written in Go and released under the MIT license. See the source, releases and issues on GitHub. Related tools: Foxmayn Frappe Manager for benches and Foxmayn Frappe Types for TypeScript.