Blurt

Security

Blurt hears what you say, so your words and your provider keys are what we protect first.

Found a security problem? Email support@teamnebula.ai privately. Please don’t post it publicly. Say what the problem lets someone do, how to reach it, and which Blurt version you saw it on (Settings shows the version).

We acknowledge reports within three working days and assess them within ten. Fixes ship in the next release; anything that exposes dictation content or credentials ships sooner. There is no bug bounty. We credit reporters in the release notes unless you would rather stay anonymous.

What Blurt protects

Agent connections

Coding agents talk to Blurt through a private, authenticated connection that only your macOS account can reach. Each agent gets its own credential and can only see its own sessions. The agent helper never reads Blurt’s database. Text Blurt reads aloud is cleaned first, so a reply cannot slip commands into the speech engine.

Paste and keyboard access

Blurt pastes your words by pressing ⌘V in the app you are using, and a small helper listens for your dictation key. That is why it asks for Accessibility and Input Monitoring. The helper never receives what you said.

How Blurt is built

What is out of scope

Someone who can already run programs as you, or who has your unlocked Mac in hand. Blurt’s data is protected by your macOS account, not against it. Reports about a local, unprivileged program reaching your dictation, keys, agent connection or paste path are very welcome.