# Kelet

> Kelet automatically investigates production failures in LLM apps and AI agents, surfaces root causes with evidence, and generates prompt patches so you can ship fixes fast.

Kelet is a production reliability platform for AI agents and LLM applications that continuously reads your traces, clusters failure patterns across thousands of sessions, and surfaces root causes backed by evidence — not guesswork. It generates targeted prompt patches and measures before/after reliability so you know the fix actually held. Teams connect their first agent in under five minutes with no infrastructure changes required.

- **Automated Root Cause Analysis**: *Kelet reads every trace so you don't have to — failure patterns surface in minutes with evidence, not gut feeling.*
- **Prompt Patch Generation**: *From root cause to a ready-to-ship prompt patch, with before/after reliability measurements to confirm the fix worked.*
- **Signal Collection**: *Collects human signals (thumbs-down, edited responses, abandoned conversations) and synthetic LLM-as-judge evaluators to guide investigation.*
- **Multi-Agent Support**: *Handles multi-agent architectures natively, with credit assignment that identifies exactly which agent in a chain caused a failure.*
- **OpenTelemetry-Native Integration**: *Fully OTEL-compliant — any instrumented agent works out of the box; also integrates with Langfuse, Mixpanel, PostHog, and more.*
- **Framework Compatibility**: *Works with LangChain, LangGraph, PydanticAI, Mastra, CrewAI, AutoGen, LlamaIndex, Haystack, Semantic Kernel, OpenAI, Anthropic, and Gemini APIs.*
- **Continuous Operation**: *Runs 24/7 on Kelet's SOC 2 certified infrastructure — ingesting traces, finding patterns, and proposing fixes without manual intervention.*
- **Private Model Fine-Tuning**: *Automatically fine-tunes a private set of models per sub-agent, trained on your traces and never shared.*
- **Quick Setup**: *Install via `pip install kelet` or `npm install kelet`, add two lines to your agent code, and traces start flowing immediately.*

## Features
- Automated root cause analysis for LLM/agent failures
- Prompt patch generation with before/after reliability measurement
- Human signal and feedback collection
- Synthetic LLM-as-judge evaluators
- Multi-agent credit assignment
- OpenTelemetry-native trace ingestion
- Session inspector with trace waterfall
- Failure pattern clustering across thousands of sessions
- Private per-agent model fine-tuning
- SOC 2 certified infrastructure
- Dashboard with agent health scores and AI-generated brief
- 5-minute integration via SDK or OTEL

## Integrations
OpenTelemetry, Langfuse, Mixpanel, OpenAI, Anthropic, LangChain, Pydantic AI, Vercel AI SDK, CrewAI, Strands, Agno, Mastra, PostHog, LangGraph, AutoGen, LlamaIndex, Haystack, Semantic Kernel, Gemini

## Platforms
WEB, API, CLI

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://kelet.ai
- Documentation: https://kelet.ai/docs
- EveryDev.ai: https://www.everydev.ai/tools/kelet