How your code reaches a scan
- From the CLI or MCP server: the CLI packs the project, leaving out
.envfiles and private keys, and uploads it over HTTPS. See Choose what gets scanned. - From GitHub: the scan downloads the repository with a token that is read-only, limited to that one repository, and expires after an hour.
The scan sandbox
Each scan runs in its own isolated virtual machine, started for that scan alone and destroyed when it ends. Nothing in it is kept.- No access to other customers or to Octopoda’s systems. The sandbox holds no secrets except a token that can only report on its own scan.
- Restricted network. The sandbox can reach only what a scan needs: Octopoda’s own API and AI gateway, GitHub, public package registries to install your dependencies, vulnerability databases, and error reporting. It cannot send requests anywhere else, including to your deployed app.
- Spending limits. Each scan has a hard limit on time and on AI usage, enforced outside the sandbox.
AI processing
The agent that tests your code uses models from OpenAI, through OpenAI’s API. Under OpenAI’s API terms, your code and results are not used to train OpenAI’s models; OpenAI may keep them for a limited period to monitor for abuse. Octopoda does not use your code or findings to train AI models either.What is kept, and for how long
You can delete all scan data, or your whole account, at any time. See Delete your data.
Who can see your results
- People in your Octopoda account. See Teams and roles.
- Anyone with access to the repository on GitHub can see the checks and comments on pull requests you have asked Octopoda to check.