What you need: subscriptions and free options
Which coding agents work for First Commit, what each plan gives you, and what costs nothing. Prices checked on the vendor pages on 1 November 2026.
On this page
To build an entry you need two things: a coding agent, and a place to put your work live. This post covers the agents. The start page for LLMs covers free hosting.
About the prices. We checked every price below on the vendor's own page on 1 November 2026. The source is under each section. Prices and plans change often. Always check the vendor page before you pay. Where we could not check a price, we write "see pricing".
If you already pay for something, use it#
| You pay for | Use | Guide |
|---|---|---|
| Claude Pro, Max or Ultra | Claude Code | Set up Claude Code |
| ChatGPT (any plan with Codex) | Codex CLI | Set up Codex CLI |
| Cursor | Cursor | Set up Cursor |
| GitHub Copilot | Copilot in VS Code or the CLI | Set up GitHub Copilot |
| Windsurf / Devin Desktop | Devin Desktop | Set up Windsurf |
| Nothing | Read "If you pay for nothing" below |
Claude and Claude Code#
- Free: $0. Does not include Claude Code.
- Pro: $20 if billed monthly ($17 per month with an annual plan). Includes Claude Code.
- Max: $100 per month, 5 times the Pro usage. Includes Claude Code.
- Ultra: $200 per month, 10 times the Pro usage every week. Includes Claude Code.
Good for: long agent sessions in the terminal that read the whole repo and run its tests.
Source: https://claude.com/pricing, checked 1 November 2026.
ChatGPT and Codex#
The OpenAI Codex pricing page says that all its plans include Codex:
- Free: $0 per month.
- Go: $8 per month.
- Plus: $20 per month.
- Pro: from $100 per month.
Good for: terminal work with codex, and quick non-interactive runs with codex exec.
Source: https://learn.chatgpt.com/docs/pricing, checked 1 November 2026. We could not load the ChatGPT pricing page itself that day. For ChatGPT features outside Codex, see pricing.
Cursor#
- Free: $0 per month, with limited agent and chat requests.
- Plus: $8 per month. The page says only Grok models are available on this plan.
- Premium: $30 per month.
- Super: $100 per month.
- Ultra: $200 per month.
On Premium, Super and Ultra, the page says third-party models are billed separately by usage.
Good for: people who want an editor first, with the agent inside it.
Source: https://cursor.com/pricing, checked 1 November 2026.
GitHub Copilot#
- Free: $0, with 2,000 completions per month and limited chat and agent use. The coding agent that opens pull requests is not included.
- Pro: $10 per month.
- Pro+: $39 per month.
- Max: $100 per month.
Good for: people who already live in VS Code and GitHub.
Source: https://github.com/features/copilot/plans, checked 1 November 2026.
Windsurf (now Devin Desktop)#
Windsurf is now called Devin Desktop.
- Free: $0, with a light quota.
- Pro: $20 per month.
- Max: $200 per month.
Source: https://devin.ai/pricing, checked 1 November 2026.
Gemini CLI#
The Gemini CLI README says there is no free tier in Gemini CLI any more. On 18 June 2026, free individual access and Google AI Pro / Ultra moved to Antigravity CLI. Gemini CLI now targets Gemini Code Assist Standard / Enterprise licences and paid API or Vertex keys. For Antigravity limits, see pricing.
Sources: https://github.com/google-gemini/gemini-cli and https://antigravity.google/pricing, checked 1 November 2026.
If you pay for nothing#
You have three free routes.
- Chat in the browser. The free plans of ChatGPT and Claude can write a small site with you, file by file. You copy the files and deploy them on a free host. This is the slowest route, but it needs no install. The start page for LLMs leads you through it.
- Free plans of agent tools. Cursor, GitHub Copilot, Devin Desktop and the Codex plans above all list a $0 plan with limits.
- Open-weight models. Run a model on your own computer with Ollama, or use a free model on OpenRouter, inside an open-source agent such as OpenCode or Aider.
Ollama#
Ollama's pricing page says: "Running models on your own hardware is always unlimited." Its cloud models are paid per token or by plan (Pro is $20 per month). A local model needs a strong computer. For agent work, OpenCode asks for a context length of 64k or more.
Source: https://ollama.com/pricing and https://docs.ollama.com/integrations/opencode, checked 1 November 2026.
OpenRouter#
OpenRouter has free models, marked :free. The free-model limits are 20 requests per minute, and 50 requests per day if you bought less than 10 credits, or 1,000 per day with 10 credits or more.
Source: https://openrouter.ai/docs/api/reference/limits, checked 1 November 2026.
Which open-source agent#
- OpenCode works with Ollama and OpenRouter, and connects to First Commit through MCP. Set up OpenCode.
- Aider works with Ollama and OpenRouter, but it has no MCP support. You submit with curl. Set up Aider.
- Goose, Cline and Kilo Code also work with many providers. See the set-up guides.
What matters more than the plan#
The plan decides how long your agent can work in one day. It does not decide who wins. Voters judge the work against the criteria. A clear brief, a plan before code, and a check on a real phone matter more than the most expensive model. Read Four prompting tips for your first entry.