Coding AI has moved on from autocomplete to tools that take on entire tasks inside a repository: hunting bugs, writing tests, carrying a change across many files. How much it helps depends heavily on how well the tool grasps the context of your project.
Our pick: Claude Code
The strongest agent tool in this test, for developers who can state a task clearly.
What to look for when choosing
- Project context: does the tool understand the whole codebase, or only the open file?
- Reviewability: do you see every change as a diff before it is applied?
- Code exposure: is your source transferred to the vendor, or even used for training?
- Team readiness: central billing, policies and audit features are mandatory in a company.
How we weight these points is set out in our test methodology. The tool overview gives you every category at a glance.
Every tool in this category
1
Claude Code
AnthropicA terminal agent that takes on whole tasks in a repository by itself.
2GitHub Copilot
GitHub / MicrosoftThe standard in many development teams: unobtrusive, integrated, manageable.
3Cursor
AnysphereAn editor that understands the whole codebase, for people who really let AI write along.
| # | Tool | Category | Overall score | Price | Best for |
|---|---|---|---|---|---|
| 1 | Coding | included in the Claude plans, from about $20 per month | Experienced developers, automation, larger rewrites | ||
| 2 | Coding | from about $10 per month | Development teams that need a centrally managed solution | ||
| 3 | Coding | from about $20 per month | Developers who write a lot of new code and iterate quickly |
Other categories
AI assistants & chatbotsAll-rounders for research, writing, analysis and everyday questions.AI image generatorsFrom a written description to a finished image, for marketing, social media and concept work.AI translation & writing toolsTools for translation, editing, knowledge management and marketing copy.AI video generators & voice AISpeech synthesis, video clips and avatars for training, advertising and social media.