Best coding assistant: 7 tools compared by price and way of working

The best coding assistant depends on how you work, because coding assistants have moved from completing a line to working as agents: they read the project, change several files, run the tests and propose the result. This guide compares seven options by what they cost and what they include, using the prices each vendor shows, checked on October 6, 2026. Dollars are given as listed; euros are given where the vendor's page shows them from Spain.

What this guide is not: a quality ranking. We have not run all of these tools on the same project, so we do not say which writes the best code. We order them by way of working and by price. For a closer look at the four most used, see Claude Code vs Cursor vs Codex vs Copilot.

Summary: seven coding assistants compared

Table: coding assistants by price and way of working (October 6, 2026)

Claude Code

Claude Code is Anthropic's coding agent. It is included in every paid Claude plan and shares its usage limits with the chat, so what you spend coding comes out of the same balance. Pro costs $20 a month in the US ($17 billed annually) and €18 a month before tax from Spain (€15 billed annually), and Max, from $100 or €90 a month, gives five or twenty times Pro's usage. For intensive sessions Anthropic says you can switch to pay-as-you-go API credits. The plan details are in Claude pricing, and the first steps in how to use Claude Code.

Codex, with ChatGPT

Codex is OpenAI's coding agent. According to its pricing page, the Free and Go plans give limited access, Plus ($20, or €23 a month from Spain) gives more Codex usage and Pro (from €103 a month from Spain) gives longer sessions. If you already pay for ChatGPT for other things, it is the option that adds no subscription. All plans are broken down in ChatGPT pricing.

GitHub Copilot

It works inside the most used editors. The free plan includes 2,000 code completions a month and access to models such as Haiku 4.5 and GPT-5 mini. Pro ($10 a month) adds unlimited completions, the cloud agent, code review, model selection and, according to the plan card, access to third-party agents such as Claude Code and Codex. Read the small print: in the comparison table on the same page, delegating tasks to those agents from the cloud agent (a preview feature) is listed as not included in Pro and included from Pro+. Pro+ ($39) adds the premium models GitHub lists, including Opus, and more usage; Max ($100) is for intensive use with agents. The prices are in dollars.

Cursor

Cursor is a full code editor with a built-in agent. Its free plan has limited agent requests. Plus costs $8 a month and includes only Grok models; Premium, at $30, gives four times Plus's usage and adds cloud agents, code reviews with Bugbot and third-party models, which are billed separately at their API list price. For teams, it starts at $30 per user per month. Cursor says its prices exclude applicable taxes.

Google Antigravity

Google includes in its Google's Pro plan ($19.99 a month in the US, €21.99 from Spain) entry-level usage limits on the agent model of Google Antigravity, its agentic development platform, alongside the rest of the plan's benefits (Gemini apps, 5 TB of storage). It makes sense if you already pay for Google One or use Gemini every day. How it compares with the others in ChatGPT vs Gemini.

Mistral

Mistral's free plan includes limited coding sessions. Pro is built for coding all day in the terminal, the editor or the web and includes $30 a month in API credits. Its pricing page shows $14.99 a month excluding taxes in dollars and, when you choose euros and Spain, €18.14 a month including taxes.

A model on your own computer

If your code cannot leave your machine, you can use a coding model locally with Ollama, for free. Memory decides: Qwen3-Coder 30B-A3B, for example, takes about 18.6 GB in Q4_K_M, so it needs a 24 GB card. Work out your case with our VRAM calculator and follow how to run an LLM locally. From the terminal, Ollama can open agents such as Claude Code or Codex connected to its models (ollama launch claude).

Which one should you choose?

• You already pay for ChatGPT Plus: try Codex before buying anything else.
• You work in VS Code or JetBrains and want the cheapest option: GitHub Copilot Pro, $10 a month.
• You want a powerful terminal agent and use Claude: Claude Code with Pro; Max if you run short.
• You prefer an editor built around the assistant: Cursor.
• Your code cannot leave your computer: a local model with Ollama, if you have enough memory.

Most have a free plan: give two of them the same real task from your project before you pay.

How we checked

Prices and what each plan includes come from each vendor's own pages (Anthropic, OpenAI, GitHub, Cursor, Google and Mistral), viewed on October 6, 2026; Cursor's again on October 29, 2026. Dollar prices do not include any conversion or tax your bank or the vendor may apply. Anthropic says its euro prices exclude tax and Mistral says its euro price includes it; OpenAI does not say on its pricing page. The size of Qwen3-Coder comes from the GGUF file published on Hugging Face.