I built fswitch: A zero-dependency CLI to switch and chat with free LLMs (DeepSeek, Gemini, Ollama)

Biya — Sat, 18 Jul 2026 15:28:09 +0000

Tired of juggling API subscriptions, setting up heavy frameworks like LangChain, or writing custom boilerplate code just to test different LLMs?

I built fswitch-cli (fswitch) - a fast, lightweight Python CLI tool that lets you switch between models, chat with LLMs, and run autonomous coding agent tasks on free-tier and local AI models with zero configuration overhead.

Supported Providers:

OpenRouter (NVIDIA Nemotron, DeepSeek R1, Gemini Flash, and more)
Google AI Studio (Gemini 2.5 Flash/Pro)
Groq Cloud
GitHub Models
Mistral AI (Codestral)
Ollama (Local/offline models)
Native Anthropic (Claude 3.5 Sonnet)

Key Features:

Interactive REPL
Type "fswitch repl" to start a chat. You can switch models mid-conversation using the "/switch " command!
Autonomous Agent with Self-Correction
Need the model to write code and fix its own bugs? fswitch has an autonomous agent mode:
fswitch agent "create a calculator api" --test "pytest"
The agent will write the code, run the tests, read any error outputs, and self-correct until the tests pass.
Speed Benchmarking
Race models against each other on a single prompt to see which provider is fastest:
fswitch benchmark --prompt "write a binary search in Rust"

Technical Highlights:

Zero heavy dependencies: Built purely on standard Python requests + rich for terminal UI formatting.
Visuals: Real-time markdown streaming with visual "Thinking..." loading animations.

Try it out!

You can install it directly from PyPI:
pip install fswitch-cli

Check out the source code, contribute, or drop a star on GitHub:
https://github.com/Biya-dev/freeswitch

I'd love to hear your feedback, feature ideas, and what models you'd like to see added next!

DEV Community: Biya

I built fswitch: A zero-dependency CLI to switch and chat with free LLMs (DeepSeek, Gemini, Ollama)