PebbleObserve Admin
The PebbleObserve section under Admin → Organisation is where you see what your organisation is doing with PebbleAI — how much it’s costing, what users are asking, what tools are being called, and how the AI is performing.
It mirrors the user-facing PebbleObserve product but at organisation scope, with admin-only views.
In this section
| Page | What you find there |
|---|---|
| Usage | Org-wide cost, model activity, key activity, MCP server activity, endpoint activity. Daily spend chart, tabs for each perspective. |
| Logs | Request logs, audit logs, deleted keys, deleted teams. Searchable, filterable, with full request detail per row. |
| Evaluations | Set up and review evaluations against production traces and curated datasets. |
How it relates to the user PebbleObserve
The user-facing PebbleObserve product (powered by Langfuse, with the upstream overview at /langfuse/overview) is for users of the platform — flow builders looking at the traces of their flows, prompt managers working on prompt versions, evaluation owners running quality checks.
The admin views here are for organisation owners who need to:
- See the bigger picture across all users
- Track spend against budgets
- Audit who did what
- Investigate incidents
- Set up evaluations that run on production data
The two are complementary. As an admin, you’ll typically use both — these admin pages for monitoring and audit, the user-facing PebbleObserve sections for deeper analysis of specific traces.
Where to start
If this is the first time you’re looking at PebbleObserve as an admin:
- Usage — get a sense of what your org is spending and which models are being used
- Logs — explore individual requests; find the high-cost or slow ones
- Evaluations — set up your first evaluation when you’re ready to systematically measure quality
See also
- PebbleObserve (user-facing) — the deep documentation on the underlying observability platform
- Login Activity — the sign-in audit (separate from PebbleObserve)