mcp server

movegent-kit

One command to set up Claude Code, Cursor, Gemini CLI, Antigravity and Codex with your team's skills, Jira and Slack.

ednaaaj/movegent-kitcommunitylistedunreviewed

Use tool

Set up

Runs on your computer. Pick your agent and operating system.

Before you start

Node.js 20 or newer

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "ednaaaj-movegent-kit": {
      "command": "npx",
      "args": [
        "-y",
        "@ednaaaj/agent-kit"
      ]
    }
  }
}

same format Cursor, GitHub Copilot CLI, Junie, JetBrains AI Assistant, Augment (Auggie CLI), Factory Droid, Kiro, Devin CLI +12

Values in angle brackets are placeholders. Restart Claude Desktop after editing its config.Claude Desktop docs

README

synced Oct 20, 2026

agent-kit

Set up your team's AI coding agents with one command.

Everyone gets the same skills, the same Jira connection and the same habits, whether they use Claude Code, Cursor, Gemini CLI, Antigravity or Codex. Setup takes a few minutes, and each person picks only what their work needs.

Ticket to merge request. The agent reads a Jira ticket, makes a branch, writes the change, opens the merge request, and then keeps fixing CI and review comments until it is ready for you to merge.
Team rules the agent follows. Frontend guardrails for React Native and design-system projects, plus review checklists.
Lower token use. Optional tools that shrink what the agent reads and writes.
Your own skills on top. Add your team's private skills from your own git repo.

Nothing about your company is built in. You enter your own Jira site, GitLab project and Slack channel once, and they stay on your machine. No passwords or API tokens are asked for or stored.

Install

# from npm
npx @ednaaaj/agent-kit init
npx @ednaaaj/agent-kit install

# or from source
git clone git@github.com:ednaaaj/movegent-kit.git agent-kit && cd agent-kit
node bin/cli.mjs init
node bin/cli.mjs install
node bin/cli.mjs doctor

Flags: --agents=claude,cursor, --bundles=workflow,guardrails, --tools=rtk,headroom, --pack=<git-url|path>, --skip-mcp, --skip-tools, --dry-run.

Your bundle choice is saved in ~/.agent-kit/config.json, so repeat installs without a terminal reuse it.

flowchart LR
    W["workflow<br/>ticket, MR, babysit"] --> I["install"]
    G["guardrails<br/>React Native + design system"] --> I
    S["savers<br/>token-saver tools"] --> I
    P["your pack<br/>(private git repo)"] --> I
    I --> A1["Claude Code"]
    I --> A2["Cursor"]
    I --> A3["Gemini / Antigravity"]
    I --> A4["Codex"]

What `init` asks

Nothing is required: press Enter to skip any question. Answers are saved to ~/.agent-kit/config.json, on your machine only.

#	Question	Example	Used for
1	Repo folder to work in	blank	optional. Blank means the git root of the folder your agent runs in, so repos can live anywhere
2	Design-system checkout	blank	optional, only the guardrails bundle uses it
3	GitLab username	`jdoe`	MR author / assignee
4	Main GitLab project path	`group/sub/repo`	`glab` API calls
5	MR reviewers	`alice,bob`	reviewers on new MRs
6	Jira site host	`yourco.atlassian.net`	Jira links; the cloudId is looked up for you
7	Slack channel name	`#team-reviews`	where MR posts go (blank skips Slack)
8	Slack channel ID	`C0123ABCD`	from the channel's details; re-asks if invalid
9	People to DM / @mention	`Alice,U0123ABCD`	one per line, blank line to finish
10	Who is always @mentioned	`Alice`	blank = everyone listed
11	Let the agent send without asking?	`n`	default `n`: it only drafts

install asks which agents you use, which bundles you want, and which extra tools. Jira login happens in your browser the first time an agent uses Jira. Details: docs/config.md.

Bundles

Bundle	Contents	Who needs it
`workflow`	`ship`, `ticket-to-mr`, `mr-babysit`, `mr-review`, `jira-transition`, plus the `mr-intake`, `mr-shipper`, `mr-reviewer` subagents	everyone who works tickets into MRs
`guardrails`	11 `mp-*` skills and 10 subagents for React Native and a design system	frontend / mobile devs; backend devs can skip it
`savers`	token-saver tool recipes (caveman, ponytail, rtk, headroom, context-mode, graphify)	anyone who wants smaller context use

Your own pack counts as one more bundle and is always installed. See docs/packs.md.

Repository layout

agent-kit/
  bin/cli.mjs            installer
  targets.json           where each agent reads skills, subagents and MCP
  bundles/
    workflow/            skills/ and agents/
    guardrails/          skills/ and agents/
    savers/tools.json    token-saver recipes
  docs/                  detailed docs
  examples/              example config and a minimal pack
  scripts/ tests/        validator and test suite

Docs

docs/architecture.md: how the pieces fit, what install does
docs/agents.md: which agent reads what, adding an agent
docs/tools.md: token-saver tools, adding a tool
docs/packs.md: private skill packs
docs/workflow.md: ticket-to-MR pipeline and guardrail skills
docs/config.md: every config key
docs/troubleshooting.md: common problems
AGENTS.md, CONTRIBUTING.md, SECURITY.md, CHANGELOG.md

License

MIT. Third-party tools are installed from their own upstream repos under their own licences.