gforge-cli

Your AI assistant knows your backlog.

Your AI gets live access to your tickets, wiki, branches, and project history. Your scripts and pipelines do too. One binary handles both.

  • AI reads and writes tickets, wiki pages, branches, and SCM diffs
  • Works with Claude Code, Cursor, and any stdio MCP-compatible AI tool
  • Shell scripts and CI/CD pipelines, without wrestling the raw API
  • macOS · Linux · Windows · zero runtime dependencies
Download Now Read the Guide →
GForge MCP
You
I’m on #65202 — pull mtutty’s last comment so we know what happened.
⏺ get_tracker_item #65202 · include_followups
AI Agent
Got ticket #65202: Add tagging support to wiki pages.
mtutty (2 hours ago): “Spent 2 hours on merge conflicts, didn’t trust the result. Dig through git history — don’t just make the tests pass.”
Found it — a dropped property was breaking the build. That branch had since merged upstream, so I rebased fresh instead — down to 2 conflicting files instead of 20. Tags saving correctly.
You
Ok, now that’s dealt with, I see tags are getting added twice when saving a wiki page. What happened? Confirm by checking the DB first and if that’s the case check the code and report back with your findings.
AI Agent
Confirmed: duplicate rows in tag_entry — a race condition between two save hooks firing on the same new page.
You
Ok, add a follow-up to that ticket with your findings and then give me a plan to address this race condition.
⏺ add_tracker_item_followup #65202
AI Agent
Followup posted. Plan: update tagsOriginal synchronously right after the first save hook fires, so the second hook sees no diff and skips the redundant save.
Download

Get gforge-cli

Single binary, zero dependencies. Download, unzip, and go.

Your Platform
macOS Intel
gforge-darwin-amd64
v26.0
SHA256: 4106c2d441a1e863…
Download Your Platform
macOS Apple Silicon
gforge-darwin-arm64
v26.0
SHA256: a8f964a22a1d014a…
Download Your Platform
Linux 64-bit
gforge-linux-amd64
v26.0
SHA256: 52038fb6be267d6a…
Download Your Platform
Windows 64-bit
gforge-windows-amd64
v26.0
SHA256: a35e8d2b3257c90d…
Download
MCP Server

Give your AI assistant full GForge access.

Run gforge mcp in your terminal once to configure. It prompts for your server, credentials, and project, then writes .mcp.json. Your AI tool of choice picks that up and launches the MCP server automatically — giving it structured, authenticated access to your tickets, wiki, branches, and more. Because the server is a local stdio process, your GForge data flows directly to whichever AI tool you configure. What leaves your network is determined by your tool choice, not GForge.

Your AI Tool
Claude Code, Cursor…
→
stdio
gforge mcp
local process
→
HTTPS
Your GForge
self-hosted or cloud

Behind your firewall

Runs against your own GForge installation. Not a SaaS proxy. Your AI sees exactly what your credentials permit — nothing more.

Ten built-in prompts

Design docs, code review, sprint status, release notes, bug investigation, and more — common workflows built in and ready to invoke from any MCP-compatible AI tool.

Auto-configured

Run gforge mcp in your terminal once. The wizard prompts for your server, credentials, and project, then writes .mcp.json. Your AI tool discovers it automatically — nothing else to configure.

Ask your AI anything about your project. It looks it up.

The MCP server gives your AI assistant the same access you have. It can read ticket details, pull design docs from notebook pages, check SCM history, and write follow-ups — all while you’re having a normal conversation.

This is a real ticket from our own backlog: a Japanese-localization project that started as “do what we did for Acme, Inc.’s German rollout, but for Japanese.” The AI read our documented translation process and the ticket, then published its own technical analysis directly to the ticket’s notebook — no copy-pasting between tools.

It’s not just a solo conversation, either. When a colleague adds their own notebook page to the same ticket — a design alternative, a set of concerns — your AI reads that too. On this ticket, our CTO’s notes on search-indexing tradeoffs became the direct input for the AI’s next deliverable: a phased implementation plan, published back to the same ticket for the team to review.

GForge MCP
You
I’m implementing #65522 — provide the Japanese translation using the process we’ve documented in the wiki page entitled “How Translations Work in GForge.” Read up, then give me your plan.
AI Agent
Fetching ticket #65522 with follow-ups.
AI Agent
Now reading wiki page “How Translations Work in GForge.”
AI Agent
I’ve read ticket #65522: “Add Japanese Translation” and the wiki page on translations. Here’s my plan: translate the frontend UI via JSON catalogs, plus a small CSS fix for Japanese character rendering. Full-text search relevance is a separate, non-blocking gap — I’d track that as its own follow-up rather than delay this ticket.
You
Go ahead and publish your plan as a notebook page on the ticket called “Analysis.”
AI Agent
Created notebook page, “Analysis,” on ticket #65522.
You
Our tech lead just read the analysis and added a follow-up. Read it and give me your thoughts.
AI Agent
Getting follow-ups for ticket #65522. Now thinking through the tech lead’s concern — how Japanese-specific nuances affect Multi-Language Indexing.
Your AI agent’s built-in skills

Ask naturally and your AI picks the right skill — these aren’t commands you type yourself. And unlike a single gforge-cli command, each skill chains multiple steps together (read the ticket, check history, write the result) behind one request.

Design & Planning
design-from-ti

Turn a ticket into a design

Interactive design session — loads ticket context, related items, and existing notebook pages before beginning.

draft-design-for-ti

Autonomous design doc

Generates a complete technical design document from a tracker item and writes it directly to the ticket’s notebook.

refine-tis

Backlog refinement

Reviews tickets, suggests acceptance criteria, flags unclear requirements, and estimates story points.

Development
implement-from-ti

Guided implementation

Reviews the design, writes the code, writes the tests — driven by the tracker item and its notebook pages.

review-merge

AI code review with context

Reviews a branch diff with full awareness of the ticket it closes. Surfaces issues the diff alone can’t reveal.

investigate-bug

Structured bug investigation

Reads the ticket, its history, and linked commits. Returns what’s known, what’s missing, and what to check next.

Project Management
sprint-status

Instant sprint intelligence

What’s in progress, blocked, and at risk — from live tracker data, in plain language.

release-notes

Release notes from your tracker

Summarizes what shipped in a milestone from closed tickets. Ready to edit and publish.

write-docs-from-ti

Draft wiki docs

Drafts or updates wiki documentation from a tracker item. Supports user guides, API references, and runbooks.

onboard-to-project

Project orientation guide

Generates a comprehensive onboarding guide tailored to a role: developer, QA, product owner, or designer.

Read the Setup Guide →
CLI

For scripts and automation, easier than the raw API.

Can’t use AI or need to write scripts for automation? GForge CLI is easier than the raw API. The same binary is a full-featured command-line client. Write shell scripts, automate CI/CD pipelines, and query GForge without wrestling with raw HTTP calls.

No API wrestling

gforge api <endpoint> makes any authenticated request — GET, POST, PUT, DELETE — with clean output you can pipe directly to jq. No curl, no tokens to manage per call.

Branches that remember everything

gforge branch embeds the ticket ID in the branch name. Every commit to that branch links to the ticket automatically — Git or SVN, nothing extra to remember.

Move tickets without switching windows

gforge ticket transitions shows you what’s valid. gforge ticket transition moves it. Two commands to see where a ticket can go and put it there — no browser required.

Deployments, automatically tracked

gforge deploy post walks a revision range, finds every open ticket those commits touch, and stamps a deployment record on each one. Supports multiple environments via --level.

Authenticate once, script forever

Credentials in ~/.config/gforge/config.json. Set GFORGE_SERVER in your CI environment and every pipeline step is authenticated. No secrets in scripts, no tokens to rotate.

Single binary, zero setup

One file, no runtime, no package manager. Drop it in your PATH and it works. Identical behavior on macOS, Linux, and Windows — the same script runs everywhere.

Quick Start

Up and running in minutes.

Download the binary, authenticate once, and start — either from your terminal or inside your AI tools.

  1. 1

    Download & Install

    Download for your platform , unzip it, move the binary onto your PATH, and make it executable (chmod +x on macOS/Linux).

  2. 2

    Authenticate

    Run gforge login --server your-gforge-server.com

  3. 3

    Start scripting

    gforge ticket list, gforge branch, gforge api — you’re ready.

  4. 4

    Connect your AI

    Run gforge mcp in your project directory. The wizard configures credentials and writes .mcp.json. Open your AI tool of choice — it launches the server automatically.

FAQ

Common questions

Things people ask before they download.

Does it work with self-hosted GForge?
Yes. Run gforge login --server your-instance.example.com to connect to any GForge installation. There is no dependency on gforge.com — the CLI talks directly to whatever instance you point it at.
Which AI clients are supported?
Any AI client that supports stdio MCP works today — including Claude Code and Cursor. HTTPS-based MCP support is coming fall 2026, which will expand compatibility to browser-based and cloud-hosted AI tools.
Does my project data leave my environment?
The MCP server runs as a local stdio process on your machine — GForge is not involved in routing or storing any AI conversations. Whether your data leaves your network depends entirely on which AI tool you configure. A self-hosted model sees nothing external; a cloud-based AI tool like Claude Code will send data to its provider. That choice, and its implications, belongs to you.
Do I need to keep a terminal window open for the MCP server?
No. Once you run gforge mcp for the initial setup — which writes .mcp.json into your project — your AI tool reads that configuration and automatically launches and manages the MCP server as a child process whenever it needs to use GForge tools. No background terminal, no daemon to start.
What’s the difference between gforge mcp and gforge api?
gforge api is for you — a human typing shell commands or writing scripts. gforge mcp is for your AI assistant — it exposes 29 structured tools that an AI can call with proper arguments to read and write GForge data, without needing to know endpoint syntax or handle authentication directly.
Does it work with both Git and SVN?
Yes. Git is the default. Set repo_type to svn via gforge config set repo_type svn and all branch commands switch to svn copy automatically. Everything else — tickets, wiki, API access, MCP — is SCM-agnostic.

GForge in your terminal. GForge in your AI.

One binary. Download, authenticate, and start — in minutes.

Download gforge-cli Read the Full Guide →