Openlit

OpenLIT is an open source platform for observing and improving AI agents and LLM applications, built on OpenTelemetry. The vendor describes it as an agent harness engineering platform, meaning it focuses on everything around the model: tools, context, prompts, memory, guardrails and feedback loops. A single SDK, available for Python and TypeScript, instruments more than 90 LLM providers and agent frameworks and emits standard OpenTelemetry traces. From those traces the platform provides agent observability with exception monitoring, cost and token tracking, latency analysis and GPU monitoring. Evaluations can use an LLM as judge, programmatic checks or human feedback. Runtime guardrails detect prompt injection, sensitive topics and off topic conversations, and can redact secrets and personal data before they are sent or traced, using either an LLM based detector or custom regex rules. Prompt Hub, Context, Rule Engine and Vault manage prompts and configuration without redeploying, and OpenGround compares models and prompts before release. It works with coding agents and frameworks such as Claude Code, Codex, the OpenAI Agents SDK, LangGraph and CrewAI. It suits engineering teams running agents in production that want vendor neutral telemetry and the option to host everything themselves. OpenLIT is free to self host under the Apache 2.0 license. The documentation site does not publish prices for any hosted offering.