Skip to content
AgentOpen-source agent

gpt-pilot

Open-source CLI agent that codes apps step by step with a developer overseeing it

Pricing
Open source
Platforms
CLI, self-hosted
GitHub stars
33.6K
Last commit
5 months ago
Language
Python
Listing updated
today

Similar agents

About gpt-pilot

gpt-pilot is an open-source agent for coding. It is available on CLI and self-hosted.

The README describes GPT Pilot as a research project exploring how far LLMs can go in producing working, production-ready applications while a developer supervises the implementation. Rather than emitting a whole codebase at once, it works through development tasks one at a time, with separate agents for specification, architecture, task planning, coding, review, troubleshooting, debugging and documentation.

You start it from a Python command line, give it an app name and description, and it writes generated code into a workspace folder. Projects can be listed, resumed from a chosen step, or deleted. It filters the codebase so that each model request only includes code relevant to the current task. The README names OpenAI, Anthropic and Groq as LLM providers, with Azure and OpenRouter reachable through the OpenAI setting.

The README says the repository is no longer maintained and carries a security notice about malicious code that was removed in June 2026. Anyone who cloned and ran it in that window is told to rotate credentials and check for listed indicators of compromise before using it again.

What it does

Step-by-step app building
Builds an app through ordered development tasks rather than generating everything in one pass.
Multi-agent roles
Splits work among specification, architect, tech lead, developer, reviewer and debugger agents.
Project resume and rollback
Lists projects and reloads them from the latest or a specified step, which discards later progress.
Context filtering
Sends each model request only the code relevant to the current task to limit context size.
Multiple LLM providers
Configurable for OpenAI, Anthropic and Groq, with Azure and OpenRouter via the OpenAI setting.
SQLite or PostgreSQL storage
Uses SQLite by default and can be pointed at PostgreSQL through config.

Good for

  • Prototyping a small application from a written description with developer review at each step
  • Studying how multi-agent LLM pipelines handle code generation
  • Generating a starting codebase that a developer then debugs and extends

Pricing

Open source

Open-source repository; the README does not state a licence.

gpt-pilot FAQ

What do I need to run it?

The README lists Python 3.9 or newer, an LLM provider key, and optionally PostgreSQL, with setup done through a virtual environment and a config.json file.

Is the repository still maintained?

No. The README states that the repo is no longer actively maintained and that the recent notice is a security cleanup, not a resumption of development.

Where does generated code go?

The README says generated code is stored in a workspace folder, inside a folder named after the app name entered at startup.

Is gpt-pilot free?

Yes. gpt-pilot is open source, so you can use and self-host it for free.

What are the best alternatives to gpt-pilot?

The closest alternatives to gpt-pilot are clippinator, gptme, devika, Agav and Docs.

Where this comes from

  • Description: written from the project's README.