Multi-model AI coding orchestrator where a coordinator routes Markdown-defined specialist tasks to Claude, Codex, Gemini, Kimi and Grok CLIs in isolated git worktrees.
the-open-engine/zeroshot
A CLI and workflow engine that runs coding agents like Claude Code, Codex and Copilot through configurable review, repair and test gates before work is accepted.
- Stars
- 2,002
- Forks
- 189
- Open issues
- 17
- Last commit
- 8 hours ago
What is it?
zeroshot turns a coding task into a reviewed change. A worker agent implements the task, then acceptance and code reviewers check the result in parallel. Rejected work goes back to a repair worker with the review feedback and is reviewed again. The agent that wrote the code cannot approve its own output, and the engine enforces the graph's review gates and retry limits.
The workflow is defined as a graph you can edit. Each node can use a different harness and model, such as Pi, Claude Code, Codex or Copilot, and you can add security reviewers, test agents or parallel checks. Graphs can be saved as profiles and reused. Runs are bounded and recorded in a SQLite ledger.
Delivery is optional: work can stay local, be pushed to a branch, prepared as a pull request, or taken through PR, CI and merge. The CLI is MIT licensed and can run locally, self-hosted via a Docker target image, or in zeroshot Cloud.
Key features
- 01Built-in software-change graph with implementation, parallel review, repair and optional delivery.
- 02Independent reviewers check the worker's output, and rejected work returns to repair.
- 03Custom graphs let you add security, test or parallel check nodes with per-node models.
- 04Saved profiles store graph and runtime settings for reuse across tasks.
- 05Optional delivery to a local branch, pull request, or PR and CI merge flow.
- 06Runs are recorded in a durable SQLite ledger and can be inspected in a browser UI.
Use cases
- Adding independent review and repair loops around Claude Code or Codex output before merging.
- Running security or test-focused reviewer agents alongside the implementation agent.
- Turning a task description into a reviewed pull request on a GitHub repository.
- Self-hosting agent workflows on your own infrastructure with the Docker target image.
Quick start
- 1
Install the CLI globally with npm
npm install -g @the-open-engine-company/zeroshot - 2
Create input.json with your task
- 3
Start a run with the software-change template
zeroshot run \ --title "Add JSON status output" \ --template software-change \ --input input.json \ --uniform-runtime-config runtime.json
Things to know
- Local runs require installing and signing in to a supported harness such as Codex, Claude Code, GitHub Copilot or Pi.
- Running the workflow needs a model ID supported by your installed harness, and the worker edits the selected folder directly.
- PR and ship delivery currently support only GitHub, and local GitHub delivery also needs the GitHub CLI.
Topics
FAQ
- What is zeroshot?
- A CLI and workflow engine that runs coding agents like Claude Code, Codex and Copilot through configurable review, repair and test gates before work is accepted. zeroshot turns a coding task into a reviewed change. A worker agent implements the task, then acceptance and code reviewers check the result in parallel. Rejected work goes back to a repair worker with the review feedback and is reviewed again. The agent that wrote the code cannot approve its own output, and the engine enforces the graph's review gates and retry limits.
- Is zeroshot free to use? What license does it have?
- zeroshot is open source under the MIT license. Check the license terms before using it in commercial products.
- Is zeroshot still maintained?
- Actively maintained. The last commit to zeroshot was on Oct 19, 2026.
- What can I use zeroshot for?
- Adding independent review and repair loops around Claude Code or Codex output before merging. Running security or test-focused reviewer agents alongside the implementation agent. Turning a task description into a reviewed pull request on a GitHub repository. Self-hosting agent workflows on your own infrastructure with the Docker target image.
- How do I install zeroshot?
- Install the CLI globally with npm: npm install -g @the-open-engine-company/zeroshot Create input.json with your task Start a run with the software-change template: zeroshot run \ --title "Add JSON status output" \ --template software-change \ --input input.json \ --uniform-runtime-config runtime.json
- Where can I get zeroshot?
- The source code and installation instructions are on GitHub: https://github.com/the-open-engine/zeroshot
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A modular runtime and orchestration system that coordinates AI coding agents through gated pipelines for tasks, review, shipping and docs across Claude Code, Codex, Cursor and more.
A TypeScript CLI and terminal UI for defining, launching and running persistent teams of Claude Code and Codex agents in tmux.
A personal context store that gives AI coding agents like Codex, Claude Code and OpenCode persistent project knowledge via a CLI, MCP server and desktop app.
A local browser-based review surface that lets you annotate agent plans, markdown, HTML and code diffs, then send feedback back to coding agents.
A prompt-based skill that makes AI coding agents write only the code a task needs, favoring reuse and minimal solutions over large implementations.