Privacy, History and Troubleshooting
Part of Monitor.
History
Section titled “History”| Plan | Days |
|---|---|
| Free | 90 days |
| Pro | 365 days (1 year) |
| Max | 365 days (1 year) |
| Ultra | 1,095 days (3 years) |
Bars share a zero baseline and use the same scale. These are hosted retention limits; local history is separate.
The app keeps Monitor usage, plan readings and machines that stop pushing for 90 days on Free, 1 year on Pro and Max, and 3 years on Ultra. Each night it removes what is older, and a push skips hours already past that; history removed this way does not come back after an upgrade. Settings on the Monitor page shows your plan’s limit, the date before which data is gone, and when your oldest data ages out. cedr monitor on your machine keeps everything your agents’ logs still hold; cedr monitor reset clears its local store, and the next run rebuilds it.
Your Agent’s Tools
Section titled “Your Agent’s Tools”Local MCP (cedr mcp) gives your agent monitor_report, monitor_limits and monitor_sources for this machine. The hosted MCP adds monitor_usage and monitor_limits for your organization.
What Leaves This Machine
Section titled “What Leaves This Machine”- Set scopeChoose exclusions, usage sources and project labels
- Collect locallyRead agent logs into the local usage store
- Approve uploadChoose the organization and approve retained history
- Upload approved dataSend retained usage counts, model and agent identifiers, plan readings, machine identity, source-health summaries and configured project labels
Prompts, response bodies and code are not uploaded. Recurring uploads need a schedule or approved status-line automation; exclusions do not erase previously collected data.
An explicit cedr monitor push, an installed Monitor schedule, or status-line uploads with auto = true can send usage to the app. Local reports alone do not enable organization uploads. The CLI’s separate anonymous usage data is on by default and has its own opt-out.
Monitor uploads usage counts, model and agent identifiers, plan readings, machine identity and source-health summaries. Project labels follow projects = "name", "hash" or "none". It does not upload prompts, response bodies, code, commands or error text. Owners and admins on team plans can view member usage; choose the destination organization deliberately.
Exclusions prevent collection from matching working directories. Changing them does not erase data already collected or uploaded. cedr monitor reset deletes the local store only; it is not a hosted deletion command. Before rebuilding a local store, configure exclusions so the next collection uses the intended scope.
Diagnose Missing or Stale Usage
Section titled “Diagnose Missing or Stale Usage”| Symptom | Next step |
|---|---|
| An agent is absent | Run cedr monitor sources; check whether it has produced logs and whether its log path differs from the default. |
| Logs are stored elsewhere | Set the agent’s path under [paths] in monitor.toml, for example claude = "~/.claude/projects". |
| Cursor has no usage | Cursor does not provide a supported local token log. An MCP connection cannot supply the missing history. |
| Claude Code limits are missing | Follow the status-line setup in Route, then allow a session to produce a reading. |
| Local totals exist but the app is stale | Run an approved cedr monitor push explicitly and inspect its result; check the key, membership, organization and timer. |
| Costs look too low | Check for unpriced models. API-equivalent cost is an estimate, not your invoice or a billing cap. |
Run cedr doctor for additional diagnostics. Share only the error code and relevant redacted details when requesting help.