Unless you choose connected mode (autter onboard --connect, or "Connected" in autter onboard), no code, prompts, or agent usage data is ever sent to Autter. Signing in with autter login alone does not turn on uploads. Autter runs entirely on your machine and writes attribution data into your local git repository and prompts to a local SQLite. autter whoami and autter sync status show "Cloud upload: off" exactly when nothing is being uploaded.
Before you finish autter onboard, Autter does not read your AI agents' session history, and plain commands such as autter status or autter doctor do not start the background service.
The only other data Autter can send is anonymous error and usage telemetry. It is off until you opt in during autter onboard or with autter telemetry on, and you can turn it off at any time by setting telemetry_oss to off. See configuration options for details.
- Prompts are stored locally on the developer's laptop. They are never shared with teammates or Autter.
AI attribution data is written to git notes and is readable by anyone with repo access:
- Model, agent, and accepted-rate percentages
- Which lines are AI-generated
- Git profile (name, email) of the person who steered each prompt
- Error & exception telemetry — off until you opt in during
autter onboard(or withautter telemetry on), in local and connected mode alike. It never includes code or prompts. You can turn it off or redirect it to your own endpoint. See configuration options.
If you opt in to a personal dashboard, the following is uploaded to Autter Cloud:
- Agent usage telemetry — cross-agent telemetry for every tool use, MCP call, skill invocation, interruption, error, and token-usage event, along with prompts and agent responses, used for your personal analytics
- Personal Agent usage data - % AI, # of Parallel agents and other stats on the dashboard are visible only to you unless you share them with others.
- SCM profile metadata from GitHub, Bitbucket, or GitLab
See our Cloud Privacy Policy for details.
Teams and Enterprise deployments store additional data in the team instance:
- Employee identity — names, emails, and GitHub/Bitbucket/GitLab team membership
- Agent usage telemetry — cross-agent telemetry for every tool use, MCP call, skill invocation, interruption, error, and token-usage event
- Full prompts uploaded to the Autter prompt store
- Best-effort stripping of secrets and PII by default
- You can apply your own filters for enhanced detection
- Write-only: prompts are saved to your team's instance, but developers cannot read unless explicitly granted permission.
- Full agent sessions can be reviewed, summarized, and made readable to developers
- SCM PR data from GitHub, Bitbucket, and GitLab
- PR metadata: description, opener, reviewer, status
- PR diffs — processed for computing % AI code, but not stored
- Error & exception telemetry — off until you opt in; you can turn it off or redirect it to your own endpoint at any time. See configuration options.
For more information, see our Trust Center.
In self-hosted deployments, all data is sent only to your team's Autter instance.