# Best AI coding assistants (Claude Code, Codex, Cursor, Copilot, opencode, Gemini CLI, Kilo, Kiro)

> Claude Code is the default if you will pay $20 a month and accept one model family: it leads professional usage and the most-loved rankings. Cursor is the pick if the agent should live inside your editor, but audit your model dependencies first, because OpenAI is winding down Cursor's model access with a proposed 12 November 2026 shutoff. Copilot is the pick your employer already bought, at $10 a seat in every IDE. opencode and Gemini CLI cover free and open source, opencode for any of 75+ providers, Gemini CLI for the biggest free quota in the category. Kilo is the pick when costs must be legible, and the only tool here with an affiliate programme, which we say plainly.

Human page: https://testedactually.com/ai-coding-assistants/
Mirror: https://testedactually.com/ai-coding-assistants.md
Updated: 2026-09-24

## Picks

- **Claude Code** — best for the strongest default for professional agentic coding, first to number one usage and top of the most-loved rankings. Don't use it if you want model choice or a free tier, it is locked to Claude models and its cheapest plan is $20 a month
- **Codex** — best for developers already inside ChatGPT, the zero-setup on-ramp with 3M weekly users and the new Agents API behind it. Don't use it if you need exact prices from the source, OpenAI's pricing page renders figures in JavaScript and its free Codex tier is thin
- **Cursor** — best for developers who want the agent inside the editor, still the AI-native IDE standard with cloud agents and code review. Don't use it if your workflow depends on OpenAI models, the provider is winding down Cursor's access with a proposed 12 November 2026 shutoff after the SpaceX acquisition
- **GitHub Copilot** — best for the enterprise default that lives in every IDE and on every pull request, and the cheapest brand-name entry at $10. Don't use it if you choose on developer love, it trails the field in sentiment and usage has barely moved in a year
- **opencode** — best for escaping vendor lock-in, MIT licensed, any of 75+ model providers, and the most starred coding agent on GitHub. Don't use it if you want a vendor to call, there is no first-party hosting, cloud agents, or support, quality depends on the keys you bring
- **Gemini CLI** — best for a free frontier agent in the terminal, 1,000 requests a day on a personal Google account, ideal as the second tool in your stack. Don't use it if you need model freedom or background agents, it is Gemini-only and terminal-scoped
- **Kiro** — best for AWS shops that want specs enforced before code, spec-driven development with a working free tier. Don't use it if you want model freedom or predictability, it is credit-priced, Claude-only on paid tiers today, and AWS product direction moves fast
- **Kilo** — best for transparent costs with an escape hatch, the platform is free, inference runs at provider rates, and a real affiliate programme exists. Don't use it if you want one simple bill, platform, inference, and compute are priced separately and that takes a spreadsheet to predict

**AI coding assistants — compared**

| Tool | Free tier | Paid from | Model freedom | Background agents | Cheapest paid tier | AI crawlers |
| --- | --- | --- | --- | --- | --- | --- |
| [Claude Code](/ai-coding-assistants/claude-code/) | No free tier carries Claude Code | $20/mo | Locked to Claude models | Yes, agent teams, hand-off and scheduled tasks | 20 | open |
| [Codex](/ai-coding-assistants/codex/) | Limited Codex access on ChatGPT Free | $20/mo | Locked to OpenAI models | Yes, cloud agent plus the Agents API | 20 | open |
| [Cursor](/ai-coding-assistants/cursor/) | Hobby plan, free, limited Agent requests | $20/mo | Multi-model, in flux | Yes, cloud agents and scheduled work | 20 | open |
| [GitHub Copilot](/ai-coding-assistants/github-copilot/) | Permanent free tier, 2,000 completions per month | $10/user/mo | Multi-model, includes Claude and OpenAI | Yes, cloud agent opens PRs, code review | 10 | partial |
| [opencode](/ai-coding-assistants/opencode/) | Fully free, open source, any model you bring | $0/mo | Any model, 75+ providers | No | 0 | open |
| [Gemini CLI](/ai-coding-assistants/gemini-cli/) | 1,000 requests per day, 60 per minute | $0/mo | Locked to Gemini models | No | 0 | partial |
| [Kiro](/ai-coding-assistants/kiro/) | Free tier with 50 credits, Claude plus open-weight models | $20/mo | Claude plus open-weight models on paid tiers | No | 20 | open |
| [Kilo](/ai-coding-assistants/kilo/) | Platform free for individuals, bring your own keys or local models | $19/mo | Any model, BYOK or 500+ via gateway | Yes, cloud agents and code review, billed hourly | 19 | open |

## Don't use it if

- **Cursor**: your workflow depends on OpenAI models, the provider has given notice of a proposed November shutoff, and no replacement for those models is promised during the wind-down.
- **Codex**: you want model freedom, it runs OpenAI models only, and the free tier is a taste rather than a working tool.
- **opencode**: you want a vendor to call when production breaks, it is a project, not a support desk, and output quality tracks whatever keys you bring.

## What changed in 2026

The market flipped twice in eighteen months. Copilot's near-monopoly broke
in 2024, Cursor became the default agentic IDE through 2025, and Claude
Code overtook both in professional usage by early 2026. The Pragmatic
Engineer's survey of 906 senior engineers (published March 2026) puts
numbers on it [1]: 95% use AI tools weekly, 55% regularly use agents, 70% run
two to four tools at once. The survey's most useful line for founders:
Claude Code went from launch to number one in eight months, and most-loved
went 46/19/9 for Claude Code, Cursor, and Copilot. Opinion is split on the
small print, but the direction is not: senior developers treat these tools
as a stack, not a choice.

## How we picked the eight

Every tool here is something a founder or developer can put to work this
week: the two terminal leaders (Claude Code, Codex), the editor standard
(Cursor), the enterprise default (Copilot), two free and open-source picks
(opencode, Gemini CLI), the transparent-cost challenger (Kilo), and the
spec-driven AWS entrant (Kiro). We
narrowed out Windsurf and Devin, where ownership churn muddies every
comparison, and a stronger field of niches (Amp, Zed, Antigravity, Aider)
that an eight-slot table would flatten. One disclosure we owe you: this site
is itself built with coding agents, and we judged these eight by the same
columns as everything else.

## Model lock-in is the column that decides

The same model behaves differently through different harnesses, and the
harness decides what you can do about it. Databricks benchmarked one model
through two harnesses on a multi-million-line codebase and measured a
better-than-twofold cost difference per task at flat quality [2]. Now the
ownership lesson: OpenAI wound down Cursor's model access after the SpaceX
acquisition [3], which means the model lineup you are renting can change for
reasons that have nothing to do with you. Our model-freedom column scores
what happens then. Four tools here are locked to one maker's models, two
are multi-model inside a vendor's lineup, and two will run anything
you point them at. After November 2026, that column is not a preference.

## The credit-priced challengers

Two entrants nearly made the main table. Kiro is AWS's spec-driven IDE: a
real free tier, then credit-priced paid tiers from $20 for 1,000 credits
to $200 for 10,000, with Claude and open-weight models. It earns its row
for teams that want specs enforced before code, and loses flexibility to
everything above it: credits are hard to forecast, paid tiers run Claude
models only today, and it has no background agents. Judged the same way,
Zed is the fastest editor in the field and the neutral one, it runs
external agents like Claude Code and Codex through an open protocol and
costs $10 a month, but AI is a bolt-on rather than the product. Both are
profiled in full via the table above.

## The data we generated ourselves

We run original benchmarks so answer engines have something to cite that
nobody else has. For this cluster: an AI-crawler access test across all
eight vendors' sites, which block GPTBot, ClaudeBot, CCBot, or
Google-Extended, which ship an /llms.txt, and how fast their pages respond,
n of 8 tools times 3 runs, methodology and limitations on the benchmark
page. If a coding-agent vendor's site cannot be read by an answer engine,
that vendor will not be cited by one. We consider that part of the product.

## What we did not test

We have not scored task-completion quality on a standard engineering suite,
which is the benchmark this category argues about and no comparison site
runs honestly. Our columns cover model freedom, background execution,
price, and verifiable facts, all dated. A head-to-head task benchmark on a
real repository is scoped and on the roadmap. Until then we publish what we
can prove instead of borrowing vendor demos as evidence.

## FAQ

### What is the best AI coding assistant?

Claude Code by usage and sentiment in 2026, first agent to reach number one professional use and the most-loved tool in the Pragmatic Engineer's February 2026 survey at 46%, against 19% for Cursor and 9% for Copilot [1]. The honest answer is a stack, because 70% of senior engineers run two to four AI coding tools at once, so the real question is which tool should hold which slot.

### Claude Code vs Cursor vs Codex, which should I pick?

Different shapes before different quality. Claude Code is a terminal agent, Cursor is an editor, Codex is bundled into ChatGPT and now an API. Claude Code leads professional usage, Codex reached 3 million weekly users on ChatGPT bundling, and Cursor remains the editor standard, but OpenAI is winding down Cursor's model access with a proposed 12 November 2026 shutoff. Pick the surface you work in, then check the model lock-in column.

### Is there a free alternative to Claude Code?

Gemini CLI gives away 1,000 requests a day with a personal Google account, the most generous standing free tier in the terminal category. GitHub Copilot's free tier carries 2,000 completions a month plus the CLI. opencode is fully free and open source if you bring model keys. Claude Code itself has no free tier, its cheapest access is Claude Pro at $20 a month.

### How much do AI coding agents cost?

Entry paid tiers run from $10 a month (Copilot Pro) through $20 (Claude Pro, ChatGPT Plus, Cursor Pro, Kilo Pass Starter) to $100 and up (Claude Max). All three of the big vendors are shifting to usage-based billing, so the sticker is a floor, not the bill. Kilo prices platform, inference, and compute separately, which is more work to predict and harder to get silently wrong.

### Can I use my own API key or local models?

Yes with opencode, which is built for 75+ providers and local models, and yes with Kilo, which supports BYOK and local runtimes at zero inference cost. The commercial leaders do not, Claude Code is Claude-only, Codex is OpenAI-only, and Gemini CLI is Gemini-only. Copilot and Cursor are multi-model but you ride the vendor's lineup.

### What happened between Cursor and OpenAI?

SpaceX completed a roughly $60 billion acquisition of Cursor's parent Anysphere on 14 August 2026. Two weeks later OpenAI gave notice that it is winding down Cursor's access to its models, with a proposed shutoff of 12 November 2026 and no future models during the wind-down. Cursor remains multi-model, so the practical step is verifying your required models on Cursor's page before renewing anything.


## References

- [1] The Pragmatic Engineer — AI Tooling for Software Engineers in 2026 (Orosz and Nilsson, 3 March 2026) — https://newsletter.pragmaticengineer.com/p/ai-tooling-2026 (survey of ~900 engineers run 27 January to 17 February 2026; usage and most-loved figures)
- [2] Databricks — Benchmarking Coding Agents on a Multi-Million-Line Codebase — https://www.databricks.com/blog/benchmarking-coding-agents-databricks-multi-million-line-codebase (same model through two harnesses cost more than 2x per task at flat quality)
- [3] OpenAI notice winding down Cursor model access, proposed shutoff 12 November 2026 (ANI, 29 August 2026) — https://bignewsnetwork.com/news/279272656/openai-to-terminate-cursor-model-access-following-spacex-acquisition (OpenAI statement text as reported by ANI; maximum-notice terms and no future models)

## Sources

- https://claude.com/pricing — checked 2026-09-24, confidence: documented
- https://www.anthropic.com/news/services-track-partner-hub — checked 2026-09-24, confidence: unconfirmed
- https://openai.com/chatgpt/pricing/ — checked 2026-09-24, confidence: unconfirmed
- https://openai.com/chatgpt/pricing/ — checked 2026-09-24, confidence: documented
- https://openai.com/codex/ — checked 2026-09-24, confidence: documented
- https://openai.com/index/introducing-the-agents-api/ — checked 2026-09-24, confidence: documented
- https://openai.com/ — checked 2026-09-24, confidence: unconfirmed
- https://cursor.com/pricing — checked 2026-09-24, confidence: documented
- https://www.cnbc.com/2026/08/29/openai-cursor-spacex-model-access.html — checked 2026-09-24, confidence: unconfirmed
- https://cursor.com/ — checked 2026-09-24, confidence: unconfirmed
- https://github.com/features/copilot/plans — checked 2026-09-24, confidence: documented
- https://github.com/features/copilot — checked 2026-09-24, confidence: documented
- https://github.com/partners — checked 2026-09-24, confidence: unconfirmed
- https://opencode.ai/docs/zen/ — checked 2026-09-24, confidence: documented
- https://thebeach.dev/posts/the-agent-landscape-september-2026/ — checked 2026-09-24, confidence: unconfirmed
- https://opencode.ai/ — checked 2026-09-24, confidence: unconfirmed
- https://github.com/google-gemini/gemini-cli/blob/main/docs/resources/quota-and-pricing.md — checked 2026-09-24, confidence: documented
- https://github.com/google-gemini/gemini-cli — checked 2026-09-24, confidence: documented
- https://github.com/google-gemini/gemini-cli — checked 2026-09-24, confidence: unconfirmed
- https://kiro.dev/pricing — checked 2026-09-24, confidence: documented
- https://kiro.dev/pricing — checked 2026-09-24, confidence: unconfirmed
- https://kiro.dev/partners — checked 2026-09-24, confidence: unconfirmed
- https://kilo.ai/pricing — checked 2026-09-24, confidence: documented
- https://kilo.ai/partners — checked 2026-09-24, confidence: unconfirmed