Are these actually free?

The software is; the model may not be. All of these are bring-your-own-model: the software is open source and takes nothing from you, but a cloud API bills tokens directly to the model provider, and a complex task consumes a real amount. For genuinely zero cost, run an open-weights model locally through Ollama or LM Studio — at the price of being noticeably weaker than frontier cloud models on large refactoring work.

If you work in tmux, Neovim and over ssh, take OpenCode — the terminal is its home. If you write code in VS Code all day, take Cline, which shows every change as a diff for approval. If you want something resident in the background that you command from a chat app and that handles non-coding chores too, take OpenClaw. If you want AI agents driving a real terminal session from outside, take Unterm.

Why do the agents lack the offline badge while Ollama and LM Studio have it?

Because whether an agent is offline depends on where your model lives. The software runs without a network, but a cloud API obviously does not, and that is decided by your configuration rather than by the software — so OpenCode, DeepSeek Harness, Cline and OpenClaw carry no offline badge. Ollama and LM Studio are model runners: once the weights are downloaded, inference happens entirely on your machine and keeps producing results with the network unplugged, so for them the badge is literal.

← Back to collections

Free Software Collection

Free Open-Source AI Coding Agents

This category is dominated by subscriptions: pay monthly, the vendor picks the model, and your code passes through the vendor's servers. The open-source group has a different structure — the software is free and you attach the model. A cloud API is billed per token by the model provider; a local model served by Ollama or LM Studio costs nothing and keeps the code on your machine. This collection sorts them by where they run (terminal, editor, background daemon) and by licence. Star counts are as of September 2026 and can be checked against each repository.

16 curated recommendations

Selection Criteria

The software itself is open source and free, with no paywall on core features.
The model is yours to choose, with no lock-in to one provider.
It runs on your own API key or a local model, with no forced vendor account.
Licences recorded from the actual LICENSE file.
Model-side costs are stated separately — free software does not mean a free arrangement.

Recommended Tools

Free

OpenCode

An open-source coding agent for the terminal, wired to 75-plus model providers

Open Source
No Ads
No Sign-up
Cross-platform

Free

DeepSeek Harness

DeepSeek's open-source agent runtime, where everything is a plugin

Open Source
No Ads
No Sign-up
Cross-platform

Free

Cline

An open-source AI coding assistant — you choose the model

Open Source
No Ads
No Sign-up
Cross-platform

Free

Aider

Every AI edit lands as a Git commit you can walk back

Open Source
No Ads
No Tracking
No Sign-up

Free

Continue

A Copilot replacement inside VS Code and JetBrains, with the model your choice

Open Source
No Ads
No Tracking
No Sign-up

Free

OpenClaw

An AI assistant that runs on your own machine and meets you in the chat apps you already use

Open Source
No Ads
No Tracking
No Sign-up

✦ Ours

Free

Unterm

The terminal AI agents can drive

Open Source
No Ads
No Tracking
Offline

Free

Ollama

Run large language models on your own machine, one command away

Open Source
Offline
Cross-platform
No Ads

Free

LM Studio

Run local LLMs in a desktop GUI — free for personal and commercial use

Offline
Cross-platform
No Ads
No Sign-up