Skip to content

Free tool

AI Coding Assistant Comparison (2026)

Every major AI coding tool by price, billing model, form, and features: Cursor, GitHub Copilot, Claude Code, Cline, Zed, JetBrains Junie, Replit, Amazon Q, Tabnine, Augment, Gemini Code Assist, and Google Antigravity. Filter by tool, billing, or open-source. Most prices re-read on the vendor's own page 2026-09-19; a handful still carry 2026-08-17 figures, named under the table.

33 plans

All, Any billing. 33 of 33 plans listed.

  • AiderOpen-source / BYO-key
    Free
    free/BYO-keyCLIopen-source

    Terminal pair programmer, edits + auto-commits to your git repo, repo map, 100+ languages, OSS

    Models:
    Any LLM (Claude, GPT, DeepSeek, local)
    Best for:
    Terminal-first devs wanting free, model-agnostic pair programming
  • Amazon Q DeveloperFree
    Free
    usage/creditsextension

    ~50 agentic req/mo, completions, AWS-aware, scans

    Models:
    Amazon-managed (Claude on Bedrock)
    Best for:
    AWS devs trying agentic coding free
  • ClineOpen-source / BYO-key
    Free
    free/BYO-keyextensionopen-source

    Autonomous agent (plan/act), codebase context, MCP, terminal+browser, Apache-2.0

    Models:
    30+ providers incl. local (Ollama/LM Studio)
    Best for:
    Full control + BYO key, no platform markup
  • CursorHobby (Free)
    Free
    subscriptionIDE

    Limited Tab + Agent, no card

    Models:
    Frontier models, limited (Auto mode)
    Best for:
    Trying Cursor / light use
  • GitHub CopilotFree
    Free
    usage/creditsextension

    2,000 completions/mo, ~50 chat, agent mode, CLI, MCP

    Models:
    Claude Haiku 4.5, GPT-5 mini, Gemini
    Best for:
    Casual / student use in your editor
  • Google (Antigravity)Individuals (Free)
    Free
    subscriptionIDE

    Agent-first IDE: plan/execute/verify across editor+terminal+browser; unlimited tab completions and command requests; basic weekly rate limits. Google AI Pro and Ultra raise the limits

    Models:
    Gemini 3.8/3.7/3.6 Flash, Gemini 3.1 Pro, Claude Sonnet & Opus 4.6, gpt-oss-120b
    Best for:
    Google's agent-first IDE, free with no subscription
  • JetBrains (AI / Junie)AI Free
    Free
    usage/creditsIDE

    Limited completion + chat, small Junie agent quota

    Models:
    Local + limited cloud
    Best for:
    Basic in-IDE AI for JetBrains users
  • ReplitStarter (Free)
    Free
    usage/creditsweb

    Free daily agent usage, 1 live published project, built-in database, 1 background task

    Models:
    Claude (Replit Agent) + frontier
    Best for:
    Trying Replit Agent without a card
  • ZedPersonal (Free)
    Free
    free/BYO-keyIDEopen-source

    2,000 edit predictions/mo, unlimited BYO-key AI, fast native editor, MCP

    Models:
    BYO key or external agents (Claude Code)
    Best for:
    Fast native editor with BYO-key AI
  • GitHub CopilotPro
    $10/mo
    usage/creditsextension

    Unlimited completions, agent mode, cloud agent, ~$15 credits, MCP

    Models:
    Claude, GPT-5, Gemini (picker)
    Best for:
    Individual devs in VS Code / JetBrains
  • JetBrains (AI / Junie)AI Pro
    $10/mo
    usage/creditsIDE

    Enhanced cloud AI, Junie + Claude agent, codebase context, MCP; $10/mo personal, $20/mo commercial

    Models:
    Cloud models + Junie agent
    Best for:
    JetBrains devs wanting Junie
  • ZedPro
    $10/mo
    usage/creditsIDEopen-source

    Unlimited edit predictions, $5 hosted credits, overage at API+10%

    Models:
    Zed-hosted + BYO key
    Best for:
    Hosted models without managing keys
  • Amazon Q DeveloperPro
    $19/seat
    per-seatextension

    Higher limits, Java/.NET transforms, AWS integration, security scans, MCP

    Models:
    Amazon-managed (Claude on Bedrock)
    Best for:
    AWS teams modernizing codebases
  • GitHub CopilotBusiness
    $19/seat
    per-seatextension

    Org policy, IP indemnity, agent mode, 1,900 AI credits/user (~$19), MCP

    Models:
    Claude, GPT-5, Gemini (org-managed)
    Best for:
    Teams already on GitHub
  • Google (Gemini Code Assist)Standard
    $19/seat
    per-seatextension

    Unlimited completions, chat, agent, GCP integration, codebase context; $19/user with a yearly commitment, $22.80 month-to-month

    Models:
    Gemini 2.5 / 3 family
    Best for:
    Teams building on Google Cloud
  • Anthropic (Claude Code)via Claude Pro
    $20/mo
    subscriptionCLI agent

    Agentic terminal coding, full codebase context, MCP, subagents, hooks, plan mode

    Models:
    Claude Sonnet 5 + Opus 4.8
    Best for:
    Most individual Claude Code users
  • CursorPro
    $20/mo
    subscriptionIDE

    Unlimited Tab, extended Agent limits, generous Grok limits, MCP + skills + hooks, cloud agents, Bugbot on usage-based billing

    Models:
    All frontier (Claude Opus 4.8, GPT-5.6, Gemini 3.1)
    Best for:
    Most working developers
  • ReplitCore
    $20/mo
    usage/creditsweb

    $20/mo toward premium models, plan mode, unlimited workspaces, no Replit badge; $17/mo if billed annually

    Models:
    Claude (Replit Agent) + frontier
    Best for:
    Build + ship apps from the browser
  • JetBrains (AI / Junie)AI Ultimate
    $30/mo
    usage/creditsIDE

    Full Junie + Claude agent, credit rollover; $30/mo personal, $60/mo commercial

    Models:
    GPT-5, Claude Sonnet 4.6, Gemini 2.5 Pro + full Junie
    Best for:
    Heavy JetBrains + Junie users
  • ZedBusiness
    $30/seat
    per-seatIDEopen-source

    Org AI policies, training off by default, unlimited predictions, BYOK

    Models:
    BYO key (no extra charge)
    Best for:
    Teams needing AI governance
  • GitHub CopilotPro+
    $39/mo
    usage/creditsextension

    Unlimited completions, 7,000 AI credits/mo (~$70), premium models, audit logs

    Models:
    All premium models incl. Opus
    Best for:
    Heavy users wanting all premium models
  • GitHub CopilotEnterprise
    $39/seat
    per-seatextension

    Everything in Business + 3,900 AI credits/user (~$39), GitHub.com-wide features, audit

    Models:
    Claude, GPT-5, Gemini (org-managed)
    Best for:
    Large orgs standardized on GitHub
  • TabnineCode Assistant
    $39/seat
    per-seatextension

    Codebase-grounded completions + chat, SaaS/VPC/on-prem/air-gapped, zero retention · annual only

    Models:
    Multiple LLMs + private/on-prem
    Best for:
    Privacy/compliance enterprises
  • CursorTeams
    $40/seat
    per-seatIDE

    SSO, admin, team rules, Bugbot reviews, analytics, privacy mode; a Teams Premium seat is $120/user/mo and adds 5× Standard's Agent limits

    Models:
    All frontier models
    Best for:
    Teams needing admin + shared context
  • Google (Gemini Code Assist)Enterprise
    $45/seat
    per-seatextension

    Full GCP integration, admin controls, codebase customization, security; $45/user with a yearly commitment, $54 month-to-month

    Models:
    Gemini 3 family
    Best for:
    Enterprises on GCP
  • TabnineAgentic
    $59/seat
    per-seatextension

    Agentic workflows, Context Engine, terminal CLI agent, MCP · annual only

    Models:
    Multiple LLMs (BYO + hosted)
    Best for:
    Enterprises wanting private agentic coding
  • CursorPro+
    $60/mo
    usage/creditsIDE

    3× Pro's Agent limits, higher Grok Bot limits, cloud agents, Bugbot on usage billing

    Models:
    All frontier models
    Best for:
    Heavy individual agent users
  • Anthropic (Claude Code)via Claude Max 5×
    $100/mo
    subscriptionCLI agent

    ~5× Pro usage, priority access, agentic + MCP

    Models:
    Sonnet 5 + Opus 4.8 (more Opus)
    Best for:
    Daily heavy Claude Code users
  • Augment CodeBusiness
    $100/mo
    usage/creditsextension

    Flat $100/mo, no per-seat charge, up to 50 seats; $100/mo of usage included, then LLM at provider list price + a 40% fee

    Models:
    Frontier (Claude, GPT, Gemini)
    Best for:
    Teams on large codebases (flat price)
  • GitHub CopilotMax
    $100/mo
    usage/creditsextension

    Unlimited completions, 20,000 AI credits/mo (~$200), priority access, MCP

    Models:
    All premium models, priority to newest
    Best for:
    Heaviest individual Copilot users
  • ReplitPro
    $100/mo
    usage/creditsweb

    $100/mo toward premium models, 10 parallel agents, up to 15 collaborators; $95/mo if billed annually

    Models:
    Claude (Replit Agent) + frontier
    Best for:
    Shipping several apps a month from the browser
  • Anthropic (Claude Code)via Claude Max 20×
    $200/mo
    subscriptionCLI agent

    ~20× Pro usage, highest limits, parallel agents

    Models:
    Sonnet 5 + Opus 4.8
    Best for:
    Power users running agents all day
  • CursorUltra
    $200/mo
    usage/creditsIDE

    20× Pro's Agent limits, priority new features

    Models:
    All frontier + priority to new ones
    Best for:
    All-day agentic coding

USD, individual/standard tier unless marked /seat. 22 of 33 plans were re-read on the vendor's own pricing page 2026-09-19; the rest still carry 2026-08-17 figures (Cursor, GitHub Copilot, Cline, Aider, Zed, JetBrains (AI / Junie), Replit, Amazon Q Developer, Tabnine (2 plans) and Google (Antigravity)). AI coding prices move monthly. Most “subscription” prices are now a FLOOR, not a cap: Cursor, Replit, Augment, Zed Pro, JetBrains, Copilot and others meter agent/premium-model usage against an included credit allowance, so heavy users pay overages. “Open-source” = the editor/agent license (Cline, Zed), not the AI models (all proprietary); Claude Code is source-available. Confirm on the vendor's page before buying.

Want the verdict, not just the table?

The table lays out every plan; these head-to-heads add the opinionated pick, which to choose for your workflow, and why.

Frequently asked

Do I need a separate subscription for Claude Code?

No. Claude Code has no standalone subscription; it's included on every paid Claude plan. Claude Pro ($20/mo, or $200 billed annually) covers most individuals with Sonnet 4.6 and Opus 4.8 in the terminal, IDE, web, and desktop. Max 5× ($100/mo) and Max 20× ($200/mo) raise the shared usage limits ~5× and ~20× for heavy or all-day agentic use. You can alternatively pay-as-you-go via the Anthropic API, but the subscription plans are the value path for most developers.

What changed with GitHub Copilot pricing in 2026?

On June 1, 2026, every Copilot plan moved to usage-based billing. Each plan (Free, Pro $10, Business $19/user, Pro+ $39, Enterprise $39/user, Max $100) now includes a monthly allowance of GitHub AI Credits (1 credit = $0.01) that fund premium model/agent requests; once exhausted, paid plans can buy more. Code completions and next-edit suggestions stay unlimited and don't consume credits on any paid plan. So the monthly price is now a floor: heavy agent/premium-model users can incur overages.

Which AI coding tools are actually open source vs. just free to use?

Genuinely open source: Cline (Apache-2.0 agent, BYO key), Aider (Apache-2.0 terminal pair programmer that commits to your git repo) and the Zed editor (GPL/AGPL, though Zed's hosted AI credits are a paid add-on). 'Free to use' but closed source: Cursor Hobby, GitHub Copilot Free, JetBrains AI Free, Replit's Starter tier, Amazon Q Free and Google Antigravity's preview are all $0 but proprietary. Claude Code's CLI is source-available (you can read it) but not OSI open-source, and the underlying models are closed everywhere.

What's the cheapest way to get an agentic coding tool in 2026?

For $0: Cline (open-source, BYO API key; you only pay model providers, often pennies per task), Google Antigravity (free in public preview with Gemini 3), Copilot Free, or Replit's free Starter tier. For a flat ~$20/mo with hosted models included: Cursor Pro and Claude Pro (Claude Code), or Replit Core ($20). Watch for usage/credit overages: most $20 plans meter agent and premium-model usage beyond an included allowance. Augment is no longer in this bracket, its cheapest plan is now $100/mo.

What happened to Windsurf?

Windsurf is now part of Devin, Cognition's coding agent. windsurf.com redirects to devin.ai and there is no separate Windsurf pricing page any more, so the plans this table used to list can no longer be bought under that name. The figures it carried could no longer be checked against any Windsurf page, so they have been removed rather than left to age. If you were choosing Windsurf for its agentic editor, the closest picks still on sale here are Cursor and Cline.

Is Tabnine still a good pick for individual developers?

Not really in 2026. Tabnine retired its free and individual tiers and is now enterprise/team-focused, starting at $39/user/month (Code Assistant) or $59/user/month (Agentic), annual billing only. Its differentiator is privacy and deployment flexibility (SaaS, VPC, on-prem, air-gapped) with zero code retention, so it fits compliance-sensitive enterprises, not solo developers, who are better served by Cursor, Copilot Pro, Claude Code, Cline, or Zed.

How to choose an AI coding assistant

AI coding tools differ less in the underlying model than in how they fit your workflow: how they see your codebase, how they edit, what they cost, and where your code goes. Match those to how you actually work.

Editor fit and context

Some tools live in your existing editor, some are their own; some read your whole repository, some only the open file. The more relevant context a tool can pull in, the better its suggestions, so codebase awareness often matters more than the raw model.

How it edits

Autocomplete, chat, inline edits, and multi-file agentic changes are different modes. A tool that autocompletes well may be weak at large refactors, and an agent that edits many files needs more review. Pick the mode that matches your tasks.

Cost and privacy

Pricing ranges from free tiers to per-seat subscriptions to usage-based billing, and tools differ in whether your code is sent to a server or kept local. For proprietary code, where the code goes can decide the choice.

Where this answer comes from

Last checked on Jun 8, 2026, and reviewed on a 30-day schedule. Figures come from the company that publishes them, never from another site’s summary. How we check this.

Related reading

Go deeper on how this works and what to pick.

Newsletter

Liked the tool? Get the signal.

One weekly email on the AI + hardware that actually matters, from the people who build these calculators.

Free · unsubscribe anytime · no spam.