PromptOps CLI
PromptOps keeps Claude Code, Codex and Gemini CLI side by side by project and gives a lead agent MCP to launch helpers
- 70 ★ on the repository
- Last commit
Broad access
Why this level
- The app runs several agents with access to workspaces, a terminal and Git, and the AI can resolve conflicts and push
- The explorer connects to the project database, though queries are read-only
Install
- Works in
Manual install
npx @shellonback/promptopsInstalls and opens PromptOps Desktop. Ready builds are in the repository releases.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The app runs several agents with access to workspaces, a terminal and Git, and the AI can resolve conflicts and push
- The explorer connects to the project database, though queries are read-only
This is third-party code. Review the repository files before installing.
What it does
PromptOps is a desktop app for working with several CLI agents at once. You start it with `npx @shellonback/promptops` or download it from the releases. In it you add project workspaces and open sessions with Claude Code, Codex, Gemini CLI and other installed agents, which the app finds itself. Sessions survive restarts. The lead agent gets a built-in MCP server, `promptops mcp`: through it the agent opens sibling agents, collects their reports and moves work to a new session. Claude Code and Codex are wired up automatically, and other agents do the same with `promptops ctl` in an app terminal. Each helper gets its own provider, model and working mode. Risky actions stay behind approval, and ctl never passes answers to permission prompts to agents. The app also has Git with AI commit messages and conflict resolution, a database explorer for MySQL, PostgreSQL, SQLite and MongoDB with read-only queries, a floor plan of agents by client and token accounting read from agent transcripts. The open-source edition with no account and no server is built from source yourself.
Who it is for. Freelancers and small studios running several clients and projects with Claude Code, Codex and Gemini CLI at once who want every session in one window.
Good fit when
- One agent should split work: tests, README and review go to three helpers and the reports return to the lead
- Claude should do the refactor, Gemini the bundle analysis and Codex the review, all in one session
- You run five projects for different clients and want to see where an agent finished or waits for you
- You want to ask the project database in plain language and get rows from a read-only query
- After closing the app, sessions and agents should return to their places
Not a fit when
- You need a single agent in one folder: an ordinary terminal is enough
- Work runs over SSH on a remote machine: the built-in MCP is not registered for remote sessions
- You need a prebuilt open-source edition: it is built from source yourself
- Installing Rust, CMake and Node.js 22 or newer on the machine is out
Example request
Open three agents side by side in this project: one for tests, one for the README and one for review, and collect their reportsLimitations
- The app is a Tauri desktop build: ready binaries for macOS, Windows and Linux sit in the releases, while the open-source edition is built by hand and needs Node.js 24.15 or 22.22.3 and newer, Rust and CMake.
- Teams, sync and the mobile app are part of the ready PromptOps Desktop only.
- Remote SSH sessions do not get the MCP server because the app is not reachable from the remote host.
- A session gets at most 12 live agents, and spawns and prompts are rate-limited per ten minutes.
How to disable. Remove the app the usual way for your system. The Claude Code skill installed by `promptops ctl skill install` is removed by deleting its folder from `.claude/skills`. App data lives in its app-data folder and is wiped with it.
FAQ
How do I start PromptOps?
Run `npx @shellonback/promptops` or download the app from the repository releases. Then add a workspace and open a session with an installed agent.
How does a lead agent launch helpers?
Through the `promptops mcp` server: its tools `promptops_agent_list`, `promptops_agent_spawn` and others open agents alongside. Claude Code and Codex get the server automatically, and other agents use `promptops ctl`.
Do I need a PromptOps account?
Not for the open-source edition: no account and no PromptOps server, everything stays on your computer. Teams and sync belong to the ready PromptOps Desktop.
Can an agent answer permission prompts for me?
No. An agent in the blocked state gets a refusal for input from ctl, and closing agents or answering permission dialogs through MCP is not possible.
What does the database explorer do?
It finds the project database (MySQL, PostgreSQL, SQLite, MongoDB), connects to it and turns a plain-language question into a read-only query. The agent shows the result rows.
Discussion
No questions or reviews yet.
Related
CLI toolsA prompt template store with search and variable filling, plus a script that runs Claude agents across a whole project.
A tmux-based TUI that runs Claude Code, Codex and other agents in one list: status, quick prompts, worktrees and diff review
ReevesAgents runs Claude Code, Codex, Hermes and other CLI agents side by side in tmux; an MCP lets one lead the rest.
A plugin that syncs Claude Code, Codex CLI and Antigravity on one project and lets them hand work to each other.
One runtime for Claude Code, Codex, OpenCode and more: a run counts as done only when checks the agent did not write pass
MCP server of the self-hosted EmperorClaw panel: Claude Code and Codex read company knowledge, tasks and customers in scope
README in short
The README presents the project as an environment for several CLI agents and gives a quickstart: download the app or `npx`, add a folder, start a session, ask the lead to open helpers. Features follow: Git with AI, a database in plain language, agents that run agents, provider choice, a floor plan and token accounting, then the open-source build.