Kelet

About Kelet

Kelet is root-cause analysis for LLM applications and AI agents in production. Observability tools of the tracing generation are good at showing you that a run failed and what the spans were; Kelet's premise is that the expensive part is the next step — reading a haystack of traces to work out why the agent went wrong and what to change. It ingests from the stack teams already have, with OpenTelemetry and Langfuse integrations plus support for OpenAI, Anthropic, LangChain, CrewAI, Mastra, Agno, Strands, Pydantic AI, Google ADK, the Vercel AI SDK, PostHog and Mixpanel, then continuously diagnoses failures and generates a prompt patch rather than a dashboard. Human signals and feedback collection are free on every tier, which is a deliberate choice: the labels that make failure classification useful are exactly the thing most vendors put behind the paywall, so gating them would break the product for the teams most likely to adopt it. Billing is per session, where a session is one unit of agent work — a conversation, a task run or a pipeline execution — and the free tier covers 500 a month with 15-day retention. The startup tier is $400/month but free during early access with 30 days' notice before pricing changes and grandfathering for early users, so the honest read is that its real economics are not yet proven in the wild.

Key Features

Kelet Pros & Cons

Tags