Usage data & telemetry
Two questions, answered precisely: what do we collect, and what do you get to see.
What we collect, by install type
Section titled “What we collect, by install type”| Install type | What we receive |
|---|---|
| Local installs (stdio, the default) | Nothing. Telemetry is off by default (DW_TELEMETRY_OPT_IN=false); pilot installs turn it on per the agreement (below). |
Hosted endpoint (mcp.dataworkers.io) | Usage counters only (table below). |
| Payload sampling | Off everywhere. Opt-in on Enterprise only, enabled explicitly by your admin — and even then, shared infrastructure stores only hashes. |
The counters, exactly
Section titled “The counters, exactly”When telemetry is on (hosted endpoint, or a pilot install that opted in):
| Collected | Detail |
|---|---|
| Tool-call counts | By tool name × status × harness — counts, not contents |
| Tool arguments | SHA-256 hashes only — never the raw values |
| Install-funnel events | Install started/completed, first tool call |
| Client IP | Truncated to /24 before storage |
| Retention | 13 months |
Never collected, on any surface: query results, schemas, credentials, payload bodies, PII.
Pilots
Section titled “Pilots”Pilot installs run with telemetry on — it’s how we produce the usage reports a pilot is judged on, and it’s written into the pilot agreement, not slipped in. You can opt out at any time; opt-out is honored within minutes, and the pilot continues either way.
What you get to see
Section titled “What you get to see”Ask “What’s the status of my data connectors?” any time for a live report from your own install. A self-service usage dashboard for workspace admins is not built yet — we’d rather tell you that here than let you hunt for a page that doesn’t exist. Scale customers get usage reporting through their named engineer today.