Capability · Framework — agents

gptme

gptme runs as a CLI REPL on your laptop, giving an LLM a toolbox of shell, Python, file-edit, and browser actions. It's designed to be hackable: plain Python, simple prompts, and first-class support for Anthropic, OpenAI, local Ollama, and Groq.

Framework facts

Category: agents
Language: Python
License: MIT
Repository: https://github.com/gptme/gptme

Install

pipx install gptme

Quickstart

# one-shot from the CLI
gptme 'read README.md then write a one-paragraph elevator pitch'

# interactive
gptme
> read src/app.py and suggest 3 refactors
> apply suggestion 1 and run pytest

Alternatives

Open Interpreter — similar but more notebook-flavoured
Aider — focused purely on Git-aware code editing
Continue — IDE-embedded coding agent
Claude Code — Anthropic's official terminal coding agent

Frequently asked questions

How is gptme different from Open Interpreter?

gptme leans more toward a terminal chat workflow with Git-friendly file edits, while Open Interpreter centres on an interactive code interpreter experience. Both overlap but differ in ergonomics.

Can I run gptme fully locally?

Yes. Point it at Ollama, LM Studio, or any OpenAI-compatible server with an env var (OPENAI_BASE_URL). Expect slower reasoning on small local models.

Sources

gptme — GitHub — accessed 2026-04-20
gptme — docs — accessed 2026-04-20