MentionDraft
How it works Pricing Contact
Home AI disclaimer Privacy Terms Support

AI Disclaimer

Required because MentionDraft uses a generative model to write drafts.

MentionDraft can generate inaccurate, incomplete, or off-tone drafts. Treat every draft as a suggestion. You review it, you send it. MentionDraft never posts into the original conversation.

Model

Current model: OpenAI GPT-4o-mini. We may change to another hosted model of similar class. We hold the API key — you do not paste your own key anywhere.

What is sent to the model

On each draft we may send the incoming message, up to about 40 messages from the open pane (including a sample of your recent messages there) to our server. OpenAI may receive those samples to build or refresh a style card, plus the incoming message and a related-thread subset to write the draft. Style cards can be rebuilt on a draft, not only about every 30 days. We do not backfill the whole workspace. We do not use Slack data to train models.

Retention and training

We do not use Slack data to train MentionDraft or to opt in to provider training. OpenAI processes the prompt to return a completion under their API terms. MentionDraft stores the resulting draft on our server until you email us to delete it. Uninstalling the Chrome extension stops capture in that browser; it does not by itself erase server drafts or sessions. See the Privacy Policy.

Human control

MentionDraft does not make consequential decisions (legal, medical, HR, access control). It only proposes text for you to send. You are always the sender.

MentionDraft
Privacy Terms AI Disclaimer Support

© 2026 Talakraft. All rights reserved.