Best AI Agent

Home / Coding agents

The best AI coding agents

We rated 12 AI agents that plan changes, edit code, run tests and open pull requests. Each one is scored on what it can do, what it costs, how much control you keep, how it handles your code and what developers say.

Updated 8 October 2026312 sources citedHow we rate

Ratings

  1. 1
    Highest rated

    GitHub Copilot cloud agentby GitHub (Microsoft)

    GitHub's built-in coding agents that work in your IDE or turn issues into pull requests in the cloud.

    Pricing
    $19 per seat (Business) with 1,900 AI credits; extra usage at $0.01 per credit
    Best for
    Teams whose code lives on GitHub and who want agents that work from issues and pull requests
    9.7/10OutstandingRead review
  2. 2

    Claude Codeby Anthropic

    Anthropic's agentic coding tool that reads your codebase, edits files, runs commands and opens pull requests.

    Pricing
    From $20/month (Pro); Max $100-$200; Enterprise $20/seat plus API usage
    Best for
    Developers who work in the terminal and want an agent that runs commands, tests and git for them
    9.6/10OutstandingRead review
  3. 3

    Cursorby Anysphere (SpaceX)

    An AI code editor whose agents edit, test and ship code locally or in cloud VMs.

    Pricing
    From $20/month (Pro); Teams $40 per user/month; extra usage at model API rates
    Best for
    Developers who want one agent across a desktop editor, CLI and cloud VMs
    9.3/10OutstandingRead review
  4. 4

    Factory (Droids)by Factory

    Agent-native development platform whose Droids plan, code, test, review and ship software from terminal, desktop, IDE or cloud.

    Pricing
    $20/month Pro; Teams $60/month + $40 per seat
    Best for
    Enterprises that need SOC 2 Type II, ISO 42001, zero data retention and on-prem or airgapped deployment
    9.3/10OutstandingRead review
  5. 5

    OpenAI Codexby OpenAI

    OpenAI's coding agent for the terminal, IDE, ChatGPT desktop app and cloud, included in ChatGPT plans.

    Pricing
    Included in ChatGPT: Plus $20/month, Pro from $100, Business $20-$25 per user
    Best for
    Teams on ChatGPT Business or Enterprise that want coding agents under the same workspace and admin controls
    9.2/10OutstandingRead review
  6. 6

    Kiroby Amazon Web Services

    AWS's spec-driven agentic IDE, CLI and cloud agent that turns prompts into requirements, designs and pull requests.

    Pricing
    $20/month for 1,000 credits; extra credits $0.04 each
    Best for
    AWS-centric teams that want IAM Identity Center sign-in, AWS billing and central model and MCP governance
    9.0/10OutstandingRead review
  7. 7

    Devinby Cognition

    Cognition's autonomous cloud software engineer that takes tickets, works in its own VM and prepares pull requests.

    Pricing
    $20/month Pro; usage past quota at API list prices
    Best for
    Teams with a backlog of well-scoped tickets, migrations and bug fixes to delegate
    8.6/10ExcellentRead review
  8. 8

    Devin Desktopby Cognition

    The former Windsurf editor, now a command center for local and cloud coding agents with a full IDE.

    Pricing
    $20/month Pro; usage past quota at API list prices
    Best for
    Developers who want an AI-first, VS Code-compatible editor with local and cloud agents in one place
    8.6/10ExcellentRead review
  9. 9

    Ampby Amp Frontier Corporation

    Multi-model coding agent that works in cloud sandboxes called orbs, your terminal, or your own runners.

    Pricing
    Free with your own keys; $20/month Megawatt or $200/month Gigawatt with usage included
    Best for
    Teams that want agents running in parallel in cloud sandboxes, started and checked from the web, phone, terminal or Slack
    8.0/10ExcellentRead review
  10. 10

    Antigravity CLIby Google

    Google's terminal coding agent, successor to Gemini CLI, running the same agent engine as Antigravity 2.0.

    Pricing
    Free tier; Google AI Pro $19.99/month; Ultra from $99.99/month
    Best for
    Developers already paying for Google AI Pro or Ultra who want a terminal agent included
    7.5/10Very goodRead review
  11. 11

    Clineby Cline Bot Inc.

    Free, open-source coding agent for VS Code, JetBrains, terminal, desktop and browser that runs any model you choose.

    Pricing
    Free extension; inference at cost or BYOK; ClinePass $9.99/month
    Best for
    Developers who want to bring their own API keys or local models and pay providers directly
    7.3/10Very goodRead review
  12. 12

    Julesby Google

    Google's asynchronous coding agent that works on GitHub repos in cloud VMs and opens pull requests.

    Pricing
    Free (15 tasks/day); Google AI Pro $19.99/month; Ultra from $99.99/month
    Best for
    Solo developers and open-source maintainers on GitHub who want bug fixes, tests and dependency updates done in the background
    6.4/10GoodRead review

Compare the scores

Each criterion is scored from 1 to 5. The last column is what one developer pays a month on the plan a professional would use daily, using public prices only.

ToolScoreWhat it can doPricingControlSecurity and dataCustomer evidencePricingPer developer / mo
GitHub Copilot cloud agent9.7per seat + usage$39
Claude Code9.6per seat + usage$100
Cursor9.3per seat + usage$60
Factory (Droids)9.3per seat + usage$20
OpenAI Codex9.2per seat + usage$100
Kiro9.0credits$20
Devin8.6per seat + usage$20
Devin Desktop8.6per seat + usage$20
Amp8.0usage-based$20
Antigravity CLI7.5per seat + usage$20
Cline7.3free (bring your own model key)$10
Jules6.4per seat$20

How we rate

5 criteria, each with written rules for every level. The weighted total is the score out of 10. Read the full method.

30%

What it can do

Can it take a task from plan to pull request on its own, locally and in the background?

20%

Pricing

Can you work out what you will pay before talking to sales?

20%

Control

Can a developer or a team decide what it may do, and undo it?

15%

Security and data

Is your data handled to a standard you can show an auditor?

15%

Customer evidence

Do independent reviews and named customers back up the vendor's story?

Questions

What is the best AI coding agent?

GitHub Copilot cloud agent by GitHub (Microsoft) scores highest in our ratings, at 9.7 out of 10, ahead of Claude Code at 9.6. Copilot's agents suit teams already on GitHub that want background agents tied to issues, pull requests and Actions, with strict guardrails and enterprise controls. The catch is billing: since June 2026 every plan meters AI credits per token, and heavy agent users can pay well above the seat price.

How much does an AI coding agent cost?

Using public prices only, one developer pays a month on the plan a professional would use daily, using public prices only ranges from $10 (Cline) to $100 (OpenAI Codex).

Which of them publish their prices?

GitHub Copilot cloud agent, Claude Code, Cursor, Factory (Droids), OpenAI Codex, Kiro, Devin, Devin Desktop, Amp, Antigravity CLI and Jules publish clear prices. The rest show only an entry tier, leave usage limits vague, or sell by quote.

Do they train on my code?

GitHub Copilot cloud agent, Claude Code, Cursor, Factory (Droids), OpenAI Codex, Kiro, Devin, Devin Desktop, Amp, Antigravity CLI, Cline and Jules state that your data is not used to train their models, at least on business plans. For the others, check the terms before you sign; each review notes what the vendor publishes.

How do you rate them?

Every tool is scored from 1 to 5 on each of these criteria: what it can do (30%), pricing (20%), control (20%), security and data (15%), customer evidence (15%). The weighted total gives a score out of 10. Each level has a written rule, and every score links to its sources.