One project. You and your agent.
Instrument the app, ship a release, see what happened and fix it — from the dashboard and from the IDE, on the same evidence.
Develop, ship, analyse, fix.
The product is that loop. Errors still group into issues. The issue is one signal inside the project, next to the trace, the replay and the flag you can turn off.
Develop
An agent reads the docs, asks before it creates the project, turns tracing on and checks a test event. You wire the SDK.
Ship
Environment keys, release health and feature flags. A production key is a second credential, not a setting in the SDK.
Analyse
Issues, traces, analytics, heatmaps and replay share one session and one trace.
Fix
Dashboard AI reads the issue, the event or the replay. Your agent pulls the same evidence and edits the code with you.
One project, six signals.
The Talaria mark is a ringed dot with blades fanning out of it. The dot is the project. The blades are what it collects — and every blade opens onto the others.
- Errors grouped by a server-side fingerprint.
- Traces transactions and child spans, W3C Trace Context.
- Analytics, heatmaps, replay off until you turn them on for the project.
- Flags percent rollouts you or the agent can turn down.
How the dashboard is organised.
Observe
Overview, dashboards, issues, events and replays. Fingerprints group identical crashes on the server, so the inbox is a shortlist.
Analyse
Performance (transactions, RED, dependencies, Web Vitals), product analytics (events, paths, funnels, heatmaps) and release health.
Configure
Alerts on new issues and on RED. Feature flags, segments and experiments. Project settings for tracing, analytics, heatmaps and replay.
Workspace
Organisations, projects, groups, team roles, billing, GitHub and the MCP connection your agent uses.
one.newtalaria.com/northline/storefront/issues
TimeoutError upstream request timed out
Open, resolved, ignored.
Filter by severity, environment and release. Drill into the stack when you need it. Clients never compute the fingerprint — the server does, so identical crashes stay one issue.
Issues
The request, drawn as time.
Official SDKs capture transactions and child spans — database queries and outbound HTTP — plus Web Vitals and release health. A navigation and an API request can share one traceId.
Read the performance guides →Trace 7c2e…41af
What people did, not only what broke.
The same SDK sends named events. Funnels show which step loses people. Heatmaps show clicks and scroll depth. Every drop-off can open the replay, the errors and the trace.
Analytics, heatmaps and replay stay off until you turn them on for the project.
See product analytics →Top pages
Top events
Turn the rollout down.
Ship behind a percent rollout, a segment or an experiment, and turn the flag off when a release misbehaves. Alert when an issue is new or reopened, or when error rate, p95 or throughput crosses a line you set.
The key decides the environment.
Each key is tal_live_… and is bound to development, test, staging or production. Invite teammates as admin, member or viewer. SDKs send to ingest; the dashboard, MCP and sign-in stay on the API.
- Project
- storefront
- Ingest
- https://ingest.newtalaria.com
- Public key
- tal_live_••••••••3f9a
- Environments
- production · staging · development
Open a project.
Developer is $5 per month. Instrument with an agent, or send the first event from the docs.