Roadmap
Telemetry Machine is young and moving fast. Here's what we just shipped, what we're working on today, and what comes next.
Live for everyone.
4.2.0 81% 12%eu-west-2 64% 23%/checkout 47% 19%Surface
Highlight the outliers on a heatmap and Surface shows what sets them apart from everything else. Pin a problem on the exact version, region, pod, or customer behind it, starting from any trace query. No hunches needed, and anyone on the team can read the answer.
SLOs
See at a glance which services are keeping their promises and which need work. Set a target, or let the SLO agent suggest one from your real traffic. Alerts fire on real budget burn instead of every blip, and they arrive with context: one click opens Surface on the spans that broke the SLO.
Connect this application to Telemetry Machine.
Instrumentation addedOTel
Dataset and key createdTM
Production data verifiedOK
Agent setup
Paste our prompt into your coding agent. It installs our skills, instruments your service with OpenTelemetry, creates a dataset and key, and keeps going until your first traces and logs arrive.
Read the quick start guide$ tm-cli query checkout --type traces --since 15m \
--where "status_code = 'ERROR'" \
--description "Recent checkout errors" tm-cli
tm-cli gives coding agents the same access you have: read schemas, look up field values, and query traces, logs, and metrics from the terminal. Your agent can check its own work in production instead of asking you to.
In active development. Some of it is already in early users' hands.
Notebooks
Ask questions in any order and the agent digs through your traces, logs, and metrics, then writes what it finds into a post-mortem or investigation you both edit. Every claim points to a figure, and figures stay exactly as they ran, so the evidence never shifts under the story. Edit any section yourself, or undo anything the agent changed.
Why did checkout errors spike after lunch?
Errors on POST /checkout rose at 13:42, four minutes after 4.2.0
reached eu-west-2 Fig. 1.
Dashboards
Turn any query into a widget: time series, heatmaps, histograms, top values, and tables. Split by any field, even numeric ranges like payload size, then put them side by side for your team and your agents.
Next up once the work above lands.
Investigation agents
When an alert fires, an agent gets to work. It queries your traces, logs, and metrics, runs Surface on the failing spans, and checks what changed. You get a short write-up with the evidence linked, so you start from a hypothesis instead of a blank query.
Compared failing spans in Surface24 fields
Checked deploys in the last 6 hours3 found
Read error logs from checkout214 lines
Tell us at [email protected].