lotel
look
Colourwayautodaynight

Still choosing a look for lotel. Your pick sticks on this browser.

Join the waitlist

See what your coding agents did.

One small binary takes their OpenTelemetry, then forgets it on schedule.

claude_code.interaction26 spans, 18.0 s
forgets after 3d
  1. interactionwhole prompt18.02 s
  2. model call203 tokens out3.72 s
  3. Bashls -la && …blocked317 ms
  4. model call351 tokens out2.76 s
  5. Bashls -la1.36 s
  6. Bashpython3 test_calc.pyfailed115 ms
  7. model call286 tokens out3.02 s
  8. Readcalc.py8.7 ms
  9. Readtest_calc.py5.9 ms
  10. model call388 tokens out2.57 s
  11. Editcalc.py7.3 ms
  12. Bashpython3 test_calc.py60.0 ms
  13. model call89 tokens out1.36 s
A real Claude Code session in lotel: it ran the test, watched it fail, fixed one line.

Point Claude Code at it. No collector.

Claude Code already speaks OpenTelemetry. Set these and every model call, tool run and permission decision lands in lotel.

# in the shell you run claude from
export CLAUDE_CODE_ENABLE_TELEMETRY=1
export CLAUDE_CODE_ENHANCED_TELEMETRY_BETA=1
export OTEL_LOGS_EXPORTER=otlp
export OTEL_METRICS_EXPORTER=otlp
export OTEL_TRACES_EXPORTER=otlp
export OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
export OTEL_EXPORTER_OTLP_ENDPOINT=http://127.0.0.1:4318
export OTEL_EXPORTER_OTLP_HEADERS="Authorization=Bearer <token lotel prints>"

Every session comes with a receipt.

What it cost, what it ran, what it was not allowed to do. The session above, itemised.

lotel session receipt

model
claude-haiku-5-5
model calls
5
tool calls
7
blocked by permissions
1
failed
1
lines changed
+1 −1
tokens
192,294
read from cache
162,812
wall time
18.0 s

total$0.0079

use by 3d

Telemetry with a use-by date.

lotel keeps logs 7 days, traces 3 and metrics 30, or 2 GB, whichever comes first. Then it deletes whole files, so the disk never fills.

lotel's Storage page: each table's Parquet segments laid on its retention window, and when the next one is deletedlotel's Storage page: each table's Parquet segments laid on its retention window, and when the next one is deletedlotel's Storage page: each table's Parquet segments laid on its retention window, and when the next one is deletedlotel's Storage page: each table's Parquet segments laid on its retention window, and when the next one is deletedlotel's Storage page: each table's Parquet segments laid on its retention window, and when the next one is deletedlotel's Storage page: each table's Parquet segments laid on its retention window, and when the next one is deletedlotel's Storage page: each table's Parquet segments laid on its retention window, and when the next one is deletedlotel's Storage page: each table's Parquet segments laid on its retention window, and when the next one is deleted
The Storage page. Each block is one Parquet file.

Also a whole OpenTelemetry backend.

Logs, traces, metrics, a service map, and Explore for slicing by any attribute. Your apps can send to it too.

lotel's Explore page: p95 span duration per service over the last 15 minutes, the breakdown table and facet countslotel's Explore page: p95 span duration per service over the last 15 minutes, the breakdown table and facet countslotel's Explore page: p95 span duration per service over the last 15 minutes, the breakdown table and facet countslotel's Explore page: p95 span duration per service over the last 15 minutes, the breakdown table and facet countslotel's Explore page: p95 span duration per service over the last 15 minutes, the breakdown table and facet countslotel's Explore page: p95 span duration per service over the last 15 minutes, the breakdown table and facet countslotel's Explore page: p95 span duration per service over the last 15 minutes, the breakdown table and facet countslotel's Explore page: p95 span duration per service over the last 15 minutes, the breakdown table and facet counts
Explore: p95 latency by service.
lotel's service map: api-gateway calls checkout, which calls payments; each node shows its span count, error rate, p50 and p95 latency and sparklineslotel's service map: api-gateway calls checkout, which calls payments; each node shows its span count, error rate, p50 and p95 latency and sparklineslotel's service map: api-gateway calls checkout, which calls payments; each node shows its span count, error rate, p50 and p95 latency and sparklineslotel's service map: api-gateway calls checkout, which calls payments; each node shows its span count, error rate, p50 and p95 latency and sparklineslotel's service map: api-gateway calls checkout, which calls payments; each node shows its span count, error rate, p50 and p95 latency and sparklineslotel's service map: api-gateway calls checkout, which calls payments; each node shows its span count, error rate, p50 and p95 latency and sparklineslotel's service map: api-gateway calls checkout, which calls payments; each node shows its span count, error rate, p50 and p95 latency and sparklineslotel's service map: api-gateway calls checkout, which calls payments; each node shows its span count, error rate, p50 and p95 latency and sparklines
The service map.

One binary free, or $29 a month for the team.

Self-hosted

$0

One binary on your laptop or server. Apache-2.0. Ships with v0.1; the waitlist hears first.

Hosted team

$29a month, flat

Up to 10 developers, 30 days of history, hosted in the EU.

Indicative pricing. Tell us what you would pay in the form below.

Hear when hosted lotel opens.

One email, then nothing else. Self-hosters hear when v0.1 ships too.

Team size
What would you send?
Monthly agent spend