AgentOps MonitorStart free

Help

Help center

Common questions. If you do not see yours, email [email protected] or contact us.

Contact support Read the docs

Getting started

What is AgentOps Monitor?

Observability for AI agents in production. It records every tool call, LLM call, and dollar your agent reports to it.

How do I send traces?

There is no AgentOps package to install. Post JSON to POST /api/ingest with Authorization: Bearer aom_live_.... The docs have the same request in Python, Node.js, and curl, plus the error codes and size limits.

Where do I get an API key?

Sign in and open the API keys page. Keys start with aom_live_. The full key is shown only once, when you create it, so store it then.

Sending data

What counts as a run?

One agent execution you post, identified by its external_id, with up to 1,000 spans per request. Posting the same external_id again replaces that run and its spans; it does not add a run.

Do I have to send run totals?

No. Omit tokens_in, tokens_out, and cost_usd on the run and they are summed from its spans. Send them and your values are stored as sent.

What do the ingest error codes mean?

401: missing or invalid key. 400: malformed JSON or a schema violation. 413: body over 1 MB. 429: rate limit exceeded. 402 with quota_exceeded: the Free plan's runs for the month are used; the response includes runs_used and runs_included.

Why does Python's urllib get a 403?

Requests that carry urllib's default User-Agent are refused with HTTP 403 and the text error code: 1010. Set your own User-Agent header, or use the requests library as in the docs.

Using the dashboard

How do I find failed runs?

On the Runs page in your dashboard set Status to failed and, if you want, Started within to the last hour, 24h, 7 days or 30 days, then Apply filters. The filtered view is a plain URL you can bookmark. The dashboard shows the failure rate for the last 24h and 7 days, and the command palette (Ctrl+K) has a saved view, Failed runs (last 24h). On a run's page, Trace insights names the first failed span and the Errors filter shows only failed spans.

How do I find the most expensive runs?

On the Runs page, select the Cost column header to order by cost; select it again to flip the direction. Duration, Tokens and Started order the same way. The command palette has a saved view, Most expensive runs. The dashboard's Top agents (7 days) ledger lists the five agents with the highest reported cost, each linking to that agent's runs. On a run's page, Trace insights lists the most expensive spans and the spend by tool or model.

How do I export runs or a trace?

On the Runs page, Download CSV saves the runs that match the current filters, in the current sort, up to 10,000 rows (GET /api/runs/export). On a run's page, Download trace (JSON) saves that run and its spans with their input, output and error payloads, up to 5,000 spans, with a truncated flag when a run has more (GET /api/runs/<id>/export). Both need your signed-in session, not an API key.

Are there keyboard shortcuts?

Yes, on the signed-in pages. Ctrl+K (Cmd+K on Mac) or / opens the command palette: type to filter pages and saved runs views, paste a run id to open that run, or type anything else to search runs by external id. g then d, r, k or b goes to Dashboard, Runs, API keys or Billing. ? shows the list and Esc closes it. Single keys are ignored while you are typing in a field.

What does Possible loop mean on a trace?

It is a heuristic, not a verdict. Trace insights reports a possible loop when the same span type and tool or model is called with identical input 3 or more times, either in a row or with at most 2 other spans between repeats. It shows how many times, what those calls cost and took, and links to the first one. Only the first 500 spans of a run are analysed, and the run itself is not stopped: spend caps are not built yet.

What are run_url, Retry-After and request_id?

A 200 from POST /api/ingest returns run_id, span_count and run_url, the link straight to that run's trace. A 429 carries a Retry-After header with the seconds left in the current minute (1 to 60), repeated as retry_after_seconds in the body; wait that long and retry. Every response carries an X-Request-Id header and every error body repeats it as request_id. Send your own X-Request-Id (8 to 64 letters, digits, dots, underscores or hyphens) and it is echoed back, so quote it when you contact support.

Plans and billing

How much does it cost?

Free up to 10,000 runs a month. Team is $299 a month with 500,000 runs, after a 14-day free trial. Nothing is charged until day 15. Enterprise starts at $2,000 a month. See pricing and the refund and cancellation policy.

What happens when I reach my plan's runs?

On Free, new runs are refused with 402 (quota_exceeded) once 10,000 runs are used in a calendar month, until the month resets. On Team, ingest keeps going past 500,000 runs and the excess is not charged yet.

Where do I manage my subscription?

On the Billing page in your dashboard, which opens the Stripe billing portal.

Are budget caps available?

They are not available yet. AgentOps Monitor records the cost and token counts your agent reports at ingest and shows them on every run in the dashboard. A runaway call is not stopped. Spend caps and cost alerts are not built yet. See features.

Account and data

Where is my data?

Runs, spans, and API key metadata are stored against your account. See the privacy notice for retention and request details.

How do I delete my data?

Signed-in customers can erase their own account data; the privacy notice explains how. For anything else, send a privacy request.