Mergehand

Security and data

The AI agent runs without asking for permission at each step, so safety does not rely on prompts. It relies on isolation, on the server being the one that writes to your services, and on your approval.

Where your data lives

WhatWhereWho can use it
Your codeYour code host, and full clones on your runnerBuilt and tested on your runner. The server reads what it needs through your code host's API; it has no clones, but run logs and review data can contain excerpts of your code (see below)
AI sign-in or API keyOnly on your runnerThe AI agent in the run container. The Mergehand server never receives it
Tokens for your code host and trackerMergehand database, encrypted with AES-256-GCMThe server; the interface and logs show only the last 4 characters. A run gets only the token it needs
Run data – plans, reports, review drafts, logsMergehand serverMembers of the project. Logs include the agent's tool output, review data includes diffs, and environment proposals include copies of the files they are based on – so they can contain excerpts of your code. Known tokens are masked in run output

How long we keep it

Retention of personal data is in the Privacy Policy and the Data Processing Agreement.

Where your data is hosted

Isolated runs

The server writes, you approve

Your account

Report a problem

Found a security issue? Write to security@mergehand.com.