AI Coding Tools Compared
“AI coding tool” covers three different products: inline completion (suggests code as you type), chat/agents (plan and make multi-file changes), and code models you call through an API. They are compared separately below because a great agent is not automatically a great autocomplete. Everything here comes from vendor documentation; where a claim could not be verified, it is not made.
| Tool | Pricing | Free plan · API · Open source | Platforms | Verified |
|---|---|---|---|---|
| GitHub CopilotAn AI pair programmer and coding agent that works in your editor, terminal and on GitHub itself. | Free plan + paid | Free plan | VS Code, Visual Studio, JetBrains IDEs | checked 2026-10-19 Full profile |
| Claude CodeAnthropic's agentic coding tool that reads your codebase, edits files and runs commands from the terminal, IDE, desktop or browser. | Paid only | API | macOS, Windows, Linux/WSL | checked 2026-10-19 Full profile |
| OpenAI CodexOpenAI's coding agent, included with ChatGPT plans and usable with an API key. | Free plan + paid | Free plan · API | Desktop app, CLI, IDE extension | checked 2026-10-19 Full profile |
| Devin Desktop (formerly Windsurf)The renamed Windsurf editor, rebuilt around managing local and cloud coding agents. | Free plan + paid | Free plan · API | Windows, macOS, Linux | checked 2026-10-19 Full profile |
| CursorAn AI-first code editor with local and cloud agents, now sold on a shared Cursor and Grok subscription. | Free plan + paid | Free plan · API | Windows, macOS, Linux | checked 2026-10-19 Full profile |
| TabnineA private, organisation-aware AI coding agent designed to run inside a company's own environment. | Paid only | — | IDEs, CLI | checked 2026-10-19 Full profile |
| JetBrains AIAI Assistant, the Junie agent and JetBrains Air for building software with agents in JetBrains IDEs. | Free plan + paid | Free plan | IntelliJ IDEA, PyCharm, WebStorm | checked 2026-10-19 Full profile |
| Amazon Q DeveloperAWS's AI coding assistant, whose IDE plugins reach end of support on April 30, 2027 with Kiro as the recommended replacement. | Free plan + paid | Free plan | VS Code, JetBrains IDEs, CLI | checked 2026-10-19 Full profile |
| Gemini Code AssistGoogle's IDE coding assistant, now being retired in favour of the Antigravity family of products. | Paid only | — | VS Code, JetBrains IDEs | checked 2026-10-19 Full profile |
| OpenAI APIPay-per-token developer access to OpenAI's GPT models for building apps, agents and automations. | Paid only | API | Web (platform dashboard), REST API | checked 2026-10-19 Full profile |
| Claude Developer Platform (Claude API)Developer access to Anthropic's Claude models with usage-based pricing, batch discounts and prompt caching. | Paid only | API | Web (Claude Console), REST API, Amazon Bedrock | checked 2026-10-19 Full profile |
| Gemini API and Google AI StudioGoogle's developer API for Gemini models, with a free tier in Google AI Studio and paid tiers for production. | Free plan + paid | Free plan · API | Web (Google AI Studio), REST API | checked 2026-10-19 Full profile |
| OpenRouterOne OpenAI-compatible API key for hundreds of models from many providers. | Free plan + paid | Free plan · API | Web, REST API | checked 2026-10-19 Full profile |
Pricing models and plan flags come from each vendor's own site, with the date we last checked them on the tool's profile. They change often — confirm on the official site before deciding.
How to choose
- Match the tool to how you work. Inline completion fits existing editors and fast typists. Agents fit larger refactors and unfamiliar codebases. API access matters only if you are building something yourself.
- Free tiers differ enormously. Some assistants are free for individuals with limited completions; agents are usually metered. Each profile lists the verified plan details and the date they were checked.
- Review everything an agent writes. AI-generated code can be wrong or insecure in ways that compile cleanly. Every vendor says the same. Treat agent output as a first draft from a junior colleague — reviewed, tested, never merged blind.
- Check privacy controls. Coding tools send your code to remote models unless they run locally. Enterprise tiers usually add retention controls; the profile's privacy note covers what we could verify.
We do not run hands-on benchmarks of these products. Claims in this guide come from vendor documentation and pricing pages, with the limits and dates noted — anything we could not verify is not stated.
Frequently asked questions
What is the difference between an AI coding assistant and an AI coding agent?
An assistant suggests lines or functions as you type, inside your editor. An agent takes a goal (“fix this failing test”), plans across several files, makes the edits and may run commands. Agents are more capable on large tasks and more dangerous unreviewed.
Are there good free AI coding tools?
Yes — several major assistants have free individual tiers, and open-weight code models can be run or used through free API allowances. The table lists the verified plan status of each; free here means a standing free plan, not a trial.
Can an AI tool debug my code?
Tools can explain an error and propose fixes, and agents can attempt the fix themselves. Whether the fix is correct depends on the bug. Verified claims about debugging support come from vendor documentation; treat any outcome as a proposal to test.
Where to go next
- The AI Tools Directory. Browse every tool we have verified, filterable by pricing model, free plan, API access and open-source licence.
- Model comparisons. Compare AI models side by side on context window, inputs and verified API pricing.
- The launch tracker. See what is announced but not yet released — dates appear only when a reliable source supports them.