AI analysis
How to turn on AI with your OpenAI or Anthropic key, choose models and adjust the prompt.
Capto has two AI analyses, which run with your company's own key (OpenAI or Anthropic). That way the cost stays with whoever uses it and the data does not go through a shared AI account.
| Analysis | What it does | When it runs |
|---|---|---|
| Bug analysis | Reads steps, network and console and writes the technical description, severity, steps and likely cause | Right after the report is sent |
| Code analysis | Reads the front-end and back-end code and comments on the card with the file, the line and the fix | After the card is created, if repositories are connected |
Turn it on
- step 01Open AI Analysis
In the dashboard, open the AI Analysis tab.
- step 02Paste your keys
Enter your OpenAI and/or Anthropic key and turn on the provider.
- step 03Choose the model for each analysis
For bugs, a fast model is usually enough. For code, Capto recommends Claude Sonnet.
- step 04Turn on the analyses
Enable bug analysis and, if you want, code analysis.
Available models
| Provider | Bug analysis | Code analysis |
|---|---|---|
| OpenAI | GPT-4o mini, GPT-4o | GPT-4o, GPT-4o mini |
| Anthropic | Claude Haiku 5.5, Claude Sonnet 5.5 | Claude Sonnet 5.5, Claude Opus 5.5, Claude Haiku 5.5 |
Screenshots in the analysis
Text only sends steps, events and network failures (about 500 tokens per report). Text + screenshots lets the AI see what the user saw, such as error messages on screen (about 3,000 to 8,000 tokens per report).
Prompt
You can adjust the prompt for each analysis with your team's rules, such as severity standards or business terms. If you leave it empty, Capto uses the default prompt.