agentspeed.
Live demo · read-only

A real AgentSpeed dashboard, sample data — click around.

Switch the time window, open a run, read its span trace. Nothing to install. See the API →

agentspeed.com/demo
Projects · production
Checkout assistantdemo
Success rate
92.5%
49 succeeded
Failure rate
7.5%
4 failed / timeout
Avg quality
87.6%
across scored runs
p95 latency
5.6s
p50 3.8s
Total runs
53
last 7 days
Cost
$0.57
106.1K tokens
Quality by model
avg score · last 7 days
gpt-4o87% · 21
claude-haiku-4-591% · 11
claude-sonnet-4-681% · 11
claude-opus-4-891% · 10
Runs
last 7 days
Recent runs · click a row to see its trace
StatusStartedLatencyModelTokensCostError
succeeded4m ago4.8sgpt-4o2,308$0.01
succeeded14m ago3.1sgpt-4o1,783$0.0086
succeeded2h ago3.1sclaude-sonnet-4-62,778$0.02
succeeded4h ago5.6sgpt-4o2,062$0.01
succeeded4h ago5.2sgpt-4o1,897$0.0073
failed5h ago5.0sgpt-4o2,265$0.0100RateLimitError
succeeded5h ago5.8sgpt-4o1,512$0.0097
succeeded5h ago4.4sgpt-4o2,460$0.01
succeeded7h ago5.5sclaude-haiku-4-52,248$0.0082
succeeded7h ago4.0sclaude-sonnet-4-61,454$0.0062
succeeded8h ago5.2sgpt-4o1,957$0.01
succeeded9h ago5.5sgpt-4o1,434$0.0087

Sample data for illustration. Your dashboard shows your own runs the moment you report them.

Time to first run
One POST request. Runs appear in this view the instant you report them: no SDK to install, no batching, no waiting for a pipeline.

Point it at your agents.

Make a key, send one run, and this is your dashboard. No SDK required.

Free for 10,000 events a month. Metadata-first: we never store your prompts or outputs.