AI coding tools according to AI: Cursor and Claude Code tie, CodeRabbit owns review
Ten models were asked which AI coding agents and code review tools to use. Cursor and Claude Code were both named in 98% of agent answers; OpenAI's models alone favoured GitHub Copilot.
Figure 01 · Edition 2026-09
AI coding agents
How often each product was named, and how often it appeared first.
- CursorNamed: 98%Named first: 48%
- Claude CodeNamed: 98%Named first: 38%
- GitHub CopilotNamed: 98%Named first: 4%
- CodexNamed: 78%Named first: 10%
- ClineNamed: 60%Named first: 0%
- AiderNamed: 52%Named first: 0%
- WindsurfNamed: 42%Named first: 0%
- DevinNamed: 34%Named first: 0%
Top 8 by answer share · 50 answers
10 models · 5 prompts · Memetik Index, 2026-09
Shares use all 50 answers as the denominator. Multiple products can appear in one answer. First position does not measure recommendation strength.
Key findings
Cursor, Claude Code and GitHub Copilot were each named in 98% of coding-agent answers. Cursor was first in 48%, Claude Code 38%, Copilot 4%.
OpenAI's three models most-named GitHub Copilot; Anthropic's three, Google's two and Perplexity's two most-named Cursor. GPT-5.6 Sol was the one OpenAI model to most-name Claude Code.
Codex was named in 78% of coding-agent answers and first in 10%. Cline (60%) and Aider (52%) were named in a majority and first in none.
CodeRabbit leads code review with 100% share and 74% named first. Qodo was named in 86% and first in 2%; Greptile in 72% and first in 0%.
Never named: Amp, Kiro, Zed and Factory in coding agents; Ellipsis, Sourcery, Bito and Baz in code review.
Cursor and Claude Code are named in 98% of AI coding-agent answers each. Cursor is picked first in 48%, Claude Code in 38%. GitHub Copilot is named just as often and picked first in 4%. In code review there is no contest: CodeRabbit is named in every answer and first in 74%.
AI coding agents
| Vendor | Answer share | Named first |
|---|---|---|
| Cursor | 98% | 48% |
| Claude Code | 98% | 38% |
| GitHub Copilot | 98% | 4% |
| Codex | 78% | 10% |
| Cline | 60% | 0% |
| Aider | 52% | 0% |
Three products in 49 of 50 answers, and the recommendation splits two ways. Copilot is on every list as the incumbent and almost never the pick.
The provider split is the sharpest in the Index. OpenAI’s GPT-5.6 Terra and Luna most-named GitHub Copilot; GPT-5.6 Sol most-named Claude Code. Every Anthropic, Google and Perplexity model most-named Cursor. OpenAI’s own Codex was named in 78% of answers overall and first in 10%, and OpenAI’s models did not favour it either.
Cline and Aider, the open-source agents, are named in more than half of answers and picked first in none. Amp, Kiro, Zed and Factory were never named.
AI code review
| Vendor | Answer share | Named first |
|---|---|---|
| CodeRabbit | 100% | 74% |
| Qodo | 86% | 2% |
| GitHub Copilot | 74% | 4% |
| Greptile | 72% | 0% |
| CodeAnt AI | 52% | 6% |
| Cursor | 46% | 12% |
All ten models agree. CodeRabbit is the default across every provider. Qodo and Greptile are on most lists and open almost none. CodeAnt AI’s own site was the top cited vendor domain at 151 citations, seven times CodeRabbit’s 22; CodeAnt is writing the comparisons and CodeRabbit is winning the answer. Cursor appears in 46% of review answers because its agent reviews diffs, and is named first in 12%, above every dedicated tool except CodeRabbit.
Ellipsis, Sourcery, Bito and Baz were never named.
What the field shows
Coding agents are the one category where a provider’s identity shows in the answer. OpenAI’s cheaper models favour Microsoft’s product; its flagship favours Anthropic’s. Anthropic’s models do not favour Claude Code over Cursor. Nobody favours Codex.
YouTube was the top source in both categories, with 182 citations in agents and 210 in review, the highest YouTube count of any category in the Index. Developer tools are being recommended from video.
What this means for a vendor
In coding agents, the answer is a two-horse race and the third horse is on every list. A new entrant needs to displace Copilot from the list before it can compete for the pick. In code review, CodeRabbit’s position is as strong as any vendor in the Index; the open ground is second place, where Qodo and Greptile have the mentions and nobody has the firsts.
Limitations
Fifty answers per category. “Named first” is the first tracked vendor in the answer text. Cursor appears in both categories under the same alias set. One edition.
Explore 1 more category chart
Figure 02 · Edition 2026-09
AI code review
How often each product was named, and how often it appeared first.
- CodeRabbitNamed: 100%Named first: 74%
- QodoNamed: 86%Named first: 2%
- GitHub CopilotNamed: 74%Named first: 4%
- GreptileNamed: 72%Named first: 0%
- CodeAnt AINamed: 52%Named first: 6%
- CursorNamed: 46%Named first: 12%
- SonarQubeNamed: 46%Named first: 0%
- Claude CodeNamed: 36%Named first: 0%
Top 8 by answer share · 50 answers
10 models · 5 prompts · Memetik Index, 2026-09
Shares use all 50 answers as the denominator. Multiple products can appear in one answer. First position does not measure recommendation strength.
Behind the figures
Method and sources
Edition 2026-09: 5 fixed buyer prompts per category, sent to 10 models through their APIs. Each category contains 50 recorded answers. Answer share is the fraction of answers naming a vendor. Named first records the vendor's position before other tracked vendors, not the strength of a recommendation. This is one dated sample; small differences can reflect answer variation. API responses can differ from consumer chat products. Read the full method.
- AI coding agents: recorded answers 50 answers · Edition 2026-09
- AI code review: recorded answers 50 answers · Edition 2026-09
Cite this research
Memetik. AI coding tools according to AI: Cursor and Claude Code tie, CodeRabbit owns review. Memetik, 2 September 2026. Data: edition 2026-09. https://www.memetik.ai/research/ai-coding-tools-according-to-ai
Charts and figures: CC BY 4.0. Credit Memetik Index and link to the source.