> ## Documentation Index > Fetch the complete documentation index at: https://autter.dev/docs/llms.txt > Use this file to discover all available pages before exploring further. # Data and privacy > See what Autter stores locally, in Git, and in the cloud, with controls to opt out. You choose whether the Autter CLI stays local or connects to the Autter platform. You can change this choice at any time. Placeholder diagram of Autter data moving from a developer machine to Git notes and an optional organization cloud ## Storage at a glance | Data | Local-only | Connected personal dashboard | Teams or Enterprise | | ------------------------------------------------------------ | ------------------------------- | --------------------------------------------------- | ----------------------------------------------------------------- | | Prompt records | Local SQLite | Uploaded unless `prompt_storage` is `local` | Team prompt store, with best-effort secret and PII redaction | | Line-level attribution | Local Git notes | Local Git notes | Local Git notes plus team and SCM-linked processing where enabled | | Agent and model identifiers | Git notes | Git notes and connected transcript or usage records | Git notes and team analytics | | Git name and email | Git notes | Used with connected attribution | Stored as employee identity and SCM membership | | Tool calls, MCP calls, skills, errors, interruptions, tokens | Not uploaded as cloud analytics | Uploaded for personal analytics | Uploaded for team analytics | | SCM profile and pull request metadata | Not uploaded by local-only CLI | SCM profile metadata | PR description, opener, reviewer, status, and related metadata | | Pull request diffs | Not uploaded by local-only CLI | Depends on connected product workflow | Processed to compute AI percentage; not stored | | Error and exception telemetry | Sent by default | Sent by default | Sent by default unless disabled or redirected | Autter's public CLI privacy document does not state a universal cloud retention period. Check your agreement, deployment policy, or the [Autter Trust Center](https://trust.autter.dev) for current terms. ## Local-only mode Without login, Autter says it does not send code, prompts, or agent-usage data to the Autter platform. Local-only storage includes: * Prompt records in local SQLite storage * Temporary checkpoints under `.git/ai` * Attribution in Git notes, including the agent, model, line ranges, acceptance metrics, and steering developer's Git name and email Git notes are part of the repository's Git data. Anyone who receives the notes and has repository access can read their metadata. The open source CLI sends error and exception telemetry by default. Disable it with: ```bash theme={null} autter config set telemetry_oss off ``` ## Connected personal dashboards When you opt in to a personal dashboard, Autter Cloud receives: * Agent activity such as tool calls, MCP calls, skills, interruptions, errors, and token usage * Prompts and agent responses used for personal analytics * AI percentage, parallel-agent counts, and related dashboard metrics * SCM profile metadata from GitHub, GitLab, or Bitbucket Personal agent-usage data is visible only to you unless you share it. ### Transcript upload path When a commit has a captured agent transcript, Autter: 1. Reads and normalizes the local JSON or JSONL transcript. 2. Removes detected secrets before the transcript leaves the machine. 3. Queues the transcript in local durable storage as a content-addressed object. 4. Adds a `cas:` reference to the prompt record in the authorship note. 5. Lets the background service upload queued content to the connected organization environment. This process is best-effort. Missing, empty, or invalid transcripts are skipped, and transcript failures do not block commits. ## Teams and Enterprise Team deployments can also store: * Employee names, email addresses, and SCM team membership * Full prompt records with best-effort secret and personal-data stripping * Full agent sessions for review and summarization * Pull request metadata from supported SCMs * Organization-wide attribution, token, cost, and agent-usage analytics Prompt records are write-only for developers unless an administrator grants read access. Self-hosted deployments send this data to the team's own Autter instance. ## Opt out of cloud storage To clear credentials and reselect local-only mode: ```bash theme={null} autter logout autter onboard --local --force ``` This stops future connected writes from that machine. It does not automatically delete data uploaded before you opted out. Switching to local-only mode does not delete existing organization data. Use your organization's deletion process or contact Autter support for a deletion request. ## Stay connected but keep prompts local You can keep platform authentication while preventing future prompt uploads: ```bash theme={null} autter config set prompt_storage local ``` Attribution and other enabled telemetry can still sync. Use local-only mode if you want to stop connected storage as a whole. ## Disable Autter for a repository From the repository, run: ```bash theme={null} autter config --add exclude_repositories . ``` This adds the repository's remote to your machine-level exclusion list. ## Protect sensitive data * Never place credentials or secrets in agent prompts. * Keep `prompt_storage` set to `local` for repositories that cannot upload prompts. * Review Git-note access because attribution includes developer identity. * Store CI access tokens in a secret manager. * Review organization permissions before enabling prompt access for developers. For published policy details, see the [Autter privacy policy](https://autter.dev/privacy-policy) and [Trust Center](https://trust.autter.dev).