Ir para o conteúdo

State of AI 2026

Agents & Assistants

AI tools that help throughout the development process.

Coding Agents/Assistants Experience & Sentiment

Once again Claude shows a clear lead over the competition in terms of positive sentiment, with Claude Code topping the rankings even if GitHub Copilot technically has a larger user base.

Agrupar por:

Ordenar por:

1
[pt-BR] tools.coding_agents_assistants
GitHub Copilot
67.9%
30.7%
6,970
2
[pt-BR] tools.coding_agents_assistants
Claude Code
62.9%
35.7%
6,972
3
[pt-BR] tools.coding_agents_assistants
OpenAI Codex
34.5%
55.2%
9.9%
6,968
4
[pt-BR] tools.coding_agents_assistants
JetBrains AI
13.5%
53.9%
32.1%
6,958
5
[pt-BR] tools.coding_agents_assistants
Tabnine
8.1%
26.4%
65.1%
6,964
6
[pt-BR] tools.coding_agents_assistants
Supermaven
6.3%
20.5%
72.9%
6,958
7
[pt-BR] tools.coding_agents_assistants
Replit Agent & Assistant
4%
36.3%
59.5%
6,929
8
[pt-BR] tools.coding_agents_assistants
Devin
43.7%
53.4%
6,953

Experience

  • Já usei: Respondents who have used an item.
  • Conheço: Respondents who have heard about an item, but haven't used it.
  • Nunca ouvi falar: Respondents who have never heard about an item.

Sentiment

  • Positive: Respondents who are interested in learning more about a technology; or are willing to use it again.
  • Neutral: Responents who did not indicate any sentiment about a technology.
  • Negative: Respondents who are not interested in learning more about a technology; or have used it and had a negative experience.
“Used” average:25%%
“Positive” average:15%%
“Negative” average:19%%

Coding Agents & Assistants Ratios Over Time

Claude Code and OpenAI Codex are neck-and-neck in terms of respondent satisfaction, but the Anthropic coding tool is actually used way more than its rival.
Claude Code
OpenAI Codex
Devin
Tabnine
GitHub Copilot
JetBrains AI
Supermaven
Replit Agent & Assistant

Modo:

Visualização:

Uso:

Proporção de entrevistados que usaram um item
0%
20%
40%
60%
80%
100%
2025
2026
0%
20%
40%
60%
80%
100%

Other Coding Agents/Assistants

Given its popularity, OpenCode should probably have been part of the main survey options. But it makes up for that omission by establishing a comfortable lead over the competition here, with Cursor a close second.
What other coding agents or assistants are you currently using?
Respostas livres
Multiple
0%
20%
40%
60%
80%
100%
01
OpenCode
02
Cursor
03
Pi
04
Google Gemini
05
Antigravity
06
Windsurf Editor
07
Kiro
08
T3 Code
09
Kilo
10
Cline
0%
20%
40%
60%
80%
100%
% de respostas das questões
976
13%
20
10

Number of Items

How many items in this category respondents have used.
0%
6%
12%
18%
24%
30%
1
0
1,121
2
1
1,607
3
2
2,195
4
3
1,551
5
4
595
6
5
149
7
6
32
8
7
4
9
8
4
0%
6%
12%
18%
24%
30%
% de respostas da pesquisa
1.9

Paid Usage

Echoing results from the Models section, Claude Code also comes in first when it comes which coding assistants developers are willing to open their wallets for.
Out of the agents & assistants mentioned here, which one do you or your company actually pay for?
Multiple
0%
20%
40%
60%
80%
100%
01
Claude Code
3,700
02
GitHub Copilot
2,703
03
OpenAI Codex
1,526
04
🚫 None
1,252
05
JetBrains AI
349
06
Cursor
07
Devin
83
08
Replit Agent & Assistant
51
09
OpenCode
10
Kiro
0%
20%
40%
60%
80%
100%
% de respostas das questões
6,418
88%
10

Recursos recomendados

AI Engineering Fundamentals
Scott Moss
Netflix

AI Engineering Fundamentals

Add an agentic chat interface to an existing application. Then establish a rigorous eval harness to systematically improve the agent through context engineering, advanced tool use, RAG, and reliable production feedback loops.
AI Agents Fundamentals, v2
Scott Moss
Netflix

AI Agents Fundamentals, v2

Learn the foundations of agent development like tool calling, agent loops, and and evals. Add human-in-the-loop approvals for higher-stakes operations.
Obrigado aos nossos parceiros por nos apoiarem! Saiba mais.