Everything in the project you and your agent share.
Issues, traces, analytics, heatmaps, replay and flags — and two ways to read them: dashboard AI, and MCP for the coding agent.
Errors, traces and replay
Event ingestion
Single and batch ingest for exceptions and structured logs — over HTTPS with a project API key.
Server-side fingerprinting
Deduplicate on the server so identical crashes become one issue. Clients never compute fingerprints.
How grouping worksIssue triage
Open, resolved, ignored — with counts, comments, severity and first and last seen so ownership is clear.
The triage loopSearch and filters
Filter by environment, severity and more as you dig into production without drowning in noise.
Session replay
Retained session replays on a separate meter from events. Watch what the user did, then jump to the matching issue.
From error to replayPerformance and releases
Performance
Transactions, RED metrics, DB and HTTP spans, Web Vitals and release health from official SDKs. Traces follow OpenTelemetry concepts and W3C Trace Context.
Read the performance guidesReleases
Release health for crash-free sessions, adoption and new issues, with the commit you shipped. Connect GitHub so frames resolve to source.
Dashboards
Workspace and project dashboards for volume, issues, performance, vitals, analytics, replays and billing.
Alerts
Rules for new and reopened issues plus metric thresholds. Email, with webhook and Slack on paid plans.
Product analytics
Product analytics
Named track, page, screen and identify events in the same project as your errors. Off by default until you enable it.
What product analytics isConversion funnels
Ordered event sequences per session or per person, over a window you choose. Every entrant who went no further links to their replay, errors and trace.
How funnels workClick and scroll heatmaps
Which elements people click and how far they scroll, over a masked snapshot for desktop, tablet and mobile. Filter to rage, dead or error clicks.
Privacy-first analytics
First-party storage instead of ad cookies, consent-gated opt-in, PII property keys stripped server-side, and deletion by user or visitor id.
Flags, AI and agents
Feature flags and experiments
Enable a flag, roll it out by percent, target a segment and run an experiment through to a winner. Your agent can turn a flag off or set the percent over MCP.
Dashboard AI
Investigate an issue, explain an event or analyse a replay from the record you already have. Org chat covers questions that span the project.
MCP for coding agents
The agent reads these docs, creates a development project after you confirm, sends a test event and later pulls the error, the trace and the session.
Add Talaria with an agentWorkspace, SDKs and billing
Projects, orgs and invites
Multi-project workspaces with admin, member and viewer roles — built for product teams and agencies.
API keys and security
Project-scoped keys shown once at creation. HTTPS ingest only. Collaborate without sharing credentials.
Pricing you can explain
Accepted events, retained replays and data transfer — separate meters and spending caps. Flags, dashboard AI and MCP are part of the platform.
Sampling and meteringGet started.
Developer is $5 per month and includes 5,000 accepted events and 50 retained replays.