Skip to content
Back to search

openlit

OpenLIT is the open-source agent harness engineering platform: trace, evaluate, guard, and improve everything around the model in your AI agents, on OTEL.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/openlit/openlit with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/openlit/openlit'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/openlit--openlit/badge.svg)](/listings/openlit--openlit)

HTML

<a href="/listings/openlit--openlit"><img src="/api/listings/openlit--openlit/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Python:

pip install openlit

TypeScript:

npm install openlit

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 17 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. {
  2. "name": "openlit",
  3. "owner": {
  4. "name": "OpenLit",
  5. "email": "support@openlit.io"
  6. },
  7. "plugins": [
  8. {
  9. "name": "openlit-cc",
  10. "description": "OpenTelemetry-native session, tool, and edit-decision capture for Claude Code.",
  11. "author": {
  12. "name": "OpenLit"
  13. },
  14. "source": "./plugins/claude-code"
  15. }
  16. ]
  17. }
README.md · README · 680 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <div align="center">
  2. <img src="https://github.com/openlit/.github/blob/main/profile/assets/wide-logo-no-bg.png?raw=true" alt="OpenLIT logo" width="30%">
  3. # Same Agent, Smaller Bill!
  4. **Trace, evaluate, guard, and improve AI agents and coding agents with OpenTelemetry.**
  5. Agent observability · Evals · Guardrails · Prompt management · Cost & GPU monitoring · Self-host free (Apache 2.0)
  6. <a href="https://docs.openlit.io/latest/openlit/observability/agents/overview">
  7. <img src="docs/images/readme-hero-coding-agent-trace.png" alt="OpenLIT agent harness trace view showing LLM calls, tool calls and cost" width="90%">
  8. </a>
  9. <p>
  10. <a href="https://github.com/openlit/openlit/stargazers"><b>⭐ Star</b></a>&nbsp;&nbsp;&nbsp;&nbsp;
  11. <a href="#-get-started-in-5-minutes"><b>🚀 Quickstart</b></a>&nbsp;&nbsp;&nbsp;&nbsp;
  12. <a href="https://docs.openlit.io/"><b>📚 Docs</b></a>
  13. </p>
  14. [![Documentation](https://img.shields.io/badge/docs-openlit.io-blue)](https://docs.openlit.io/)
  15. [![License](https://img.shields.io/github/license/openlit/openlit)](https://github.com/openlit/openlit/blob/main/LICENSE)
  16. [![Downloads](https://static.pepy.tech/badge/openlit/month)](https://pepy.tech/project/openlit)
  17. [![Slack](https://img.shields.io/badge/Slack-join-purple?logo=slack)](https://join.slack.com/t/openlit/shared_invite/zt-2etnfttwg-TjP_7BZXfYg84oAukY8QRQ)
  18. [![X](https://img.shields.io/badge/follow-%40openlit__io-black?logo=x)](https://twitter.com/openlit_io)
  19. **[Website](https://openlit.io) · [Documentation](https://docs.openlit.io/latest/overview) · [Quickstart](#-get-started-in-5-minutes) · [Compare](https://openlit.io/compare) · [Join Slack](https://join.slack.com/t/openlit/shared_invite/zt-2etnfttwg-TjP_7BZXfYg84oAukY8QRQ)**
  20. </div>
  21. ---
  22. ## What is OpenLIT?
  23. **OpenLIT is an open-source** intelligence layer for AI ROI. It gives teams OpenTelemetry-native tracing, evaluations, guardrails, prompt and context management, and cost and GPU monitoring for the *harness* around their AI agents and coding agents, so every agent failure can be traced, scored, and turned into a harness fix. OpenLIT is free to self-host under Apache 2.0 and works with any model, framework, or harness: Claude Code, Codex, Cursor, OpenAI Agents SDK, LangGraph, CrewAI, and 70+ more integrations.
  24. ### What is agent harness engineering?
  25. An AI agent is a **model plus a harness**. The harness is everything except the model: tools, context, prompts, memory, hooks, guardrails, and feedback loops. **Agent harness engineering** is the discipline of designing, measuring, and improving that harness so agents are reliable in production. The core loop is *run → observe → evaluate → fix the harness → verify*, and OpenLIT gives you each step:
  26. | Harness engineering step | OpenLIT feature |
  27. |---|---|
  28. | **Observe** every LLM call, tool call, MCP request, retrieval and agent step | Agent observability & OpenTelemetry LLM tracing |
  29. | **Evaluate** quality, safety and cost on real traces | Evals (LLM-as-a-judge, programmatic, human feedback) + CI gates |
  30. | **Guard** the agent at runtime | Guardrails (prompt injection, sensitive topics, topic restriction) |
  31. | **Fix** the harness without redeploying | Prompt Hub, Context, Rule Engine, Vault (secrets) |
  32. | **Compare** models and prompts before shipping | OpenGround |
  33. | **Account** for cost, latency and hardware | Cost tracking, custom pricing, GPU monitoring (NVIDIA, AMD, Intel) |
  34. A production agent can involve:
  35. ```mermaid
  36. flowchart TD
  37. U([User]) --> A[AI Agent]
  38. A --> L[LLM calls]
  39. A --> T[Tool calls]
  40. A --> R[Retrieval]
  41. A --> M[Memory]
  42. A --> S[Sub-agents]
  43. A --> P[Prompts]
  44. A --> C[Code changes]
  45. L & T & R & M & S & P & C --> E{{Evaluation}}
  46. E --> O[["Cost / Quality / Errors"]]
  47. style U fill:#F97316,stroke:#7C2D12,color:#fff
  48. style A fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  49. style E fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  50. style O fill:#F97316,stroke:#7C2D12,color:#fff
  51. ```
  52. ---
  53. ## ⚡ Get started in 5 minutes
  54. ### 1. Start OpenLIT
  55. ```bash
  56. git clone https://github.com/openlit/openlit.git
  57. cd openlit
  58. docker compose up -d
  59. ```
  60. Open:
  61. ```text
  62. http://127.0.0.1:3000
  63. ```
  64. ### 2. Install the SDK
  65. Python:
  66. ```bash
  67. pip install openlit
  68. ```
  69. TypeScript:
  70. ```bash
  71. npm install openlit
  72. ```
  73. ### 3. Instrument your application
  74. Python:
  75. ```python
  76. import openlit
  77. openlit.init()
  78. ```
  79. That's it.
  80. OpenLIT automatically instruments supported LLM providers, frameworks, vector databases, and other AI infrastructure and exports OpenTelemetry traces and metrics.
  81. ### 4. Send telemetry
  82. By default, configure the OTLP endpoint:
  83. ```bash
  84. export OTEL_EXPORTER_OTLP_ENDPOINT="http://127.0.0.1:4318"
  85. ```
  86. Or:
  87. ```python
  88. import openlit
  89. openlit.init(
  90. otlp_endpoint="http://127.0.0.1:4318"
  91. )
  92. ```
  93. Open your dashboard and start exploring your AI application's traces, metrics, costs, and performance.
  94. ---
  95. # 🤖 Coding agent observability: Claude Code, Cursor, Codex, Windsurf & OpenCode
  96. AI coding agents are powerful — but understanding what they actually did can be difficult.
  97. OpenLIT gives you an OpenTelemetry-native view of coding-agent sessions.
  98. Install the CLI:
  99. ### macOS / Linux
  100. ```bash
  101. curl -fsSL https://raw.githubusercontent.com/openlit/openlit/main/cli/scripts/install.sh | sh
  102. ```
  103. ### Windows
  104. ```powershell
  105. iwr -useb https://raw.githubusercontent.com/openlit/openlit/main/cli/scripts/install.ps1 | iex
  106. ```
  107. Configure OpenLIT:
  108. ```bash
  109. openlit configure --endpoint http://127.0.0.1:4318
  110. ```
  111. Install coding-agent instrumentation:
  112. ```bash
  113. openlit coding install --vendor=all
  114. ```
  115. Or install individual integrations:
  116. ```bash
  117. openlit coding install --vendor=cursor
  118. openlit coding install --vendor=claude-code
  119. openlit coding install --vendor=codex
  120. openlit coding install --vendor=opencode
  121. ```
  122. Check your installation:
  123. ```bash
  124. openlit doctor
  125. ```
  126. Now OpenLIT can capture:
  127. ```mermaid
  128. flowchart LR
  129. S([Coding Agent Session]) --> P[User prompt]
  130. S --> L[LLM calls]
  131. S --> T[Tool calls]
  132. T --> T1[File reads]
  133. T --> T2[File edits]
  134. T --> T3[Shell commands]
  135. T --> T4[Search]
  136. S --> SA[Sub-agent activity]
  137. S --> TU[Token usage]
  138. S --> CO[Cost]
  139. S --> CI[Code impact]
  140. style S fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  141. style T fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  142. ```
  143. Explore the resulting sessions in the **Coding Agents** dashboard.
  144. ---
  145. # 🔍 Agent observability & LLM tracing (OpenTelemetry)
  146. ## Traces
  147. Understand exactly what happened during an AI request.
  148. <p>
  149. <img src="https://img.shields.io/badge/-LLM_calls-1F2937?style=flat-square" alt="LLM calls"> <img src="https://img.shields.io/badge/-Prompts_and_responses-1F2937?style=flat-square" alt="Prompts & responses"> <img src="https://img.shields.io/badge/-Tool_calls-1F2937?style=flat-square" alt="Tool calls"> <img src="https://img.shields.io/badge/-Retrieval-1F2937?style=flat-square" alt="Retrieval"> <img src="https://img.shields.io/badge/-Embeddings-1F2937?style=flat-square" alt="Embeddings"> <img src="https://img.shields.io/badge/-Vector_DB_ops-1F2937?style=flat-square" alt="Vector DB ops"> <img src="https://img.shields.io/badge/-Agent_steps-1F2937?style=flat-square" alt="Agent steps"> <img src="https://img.shields.io/badge/-Latency-1F2937?style=flat-square" alt="Latency"> <img src="https://img.shields.io/badge/-Token_usage-1F2937?style=flat-square" alt="Token usage"> <img src="https://img.shields.io/badge/-Errors-1F2937?style=flat-square" alt="Errors">
  150. </p>
  151. All represented using OpenTelemetry.
  152. ---
  153. ## 💰 LLM cost tracking
  154. Track the cost of your AI applications across:
  155. <p>
  156. <img src="https://img.shields.io/badge/-Models-1F2937?style=flat-square" alt="Models"> <img src="https://img.shields.io/badge/-Providers-1F2937?style=flat-square" alt="Providers"> <img src="https://img.shields.io/badge/-Users-1F2937?style=flat-square" alt="Users"> <img src="https://img.shields.io/badge/-Sessions-1F2937?style=flat-square" alt="Sessions"> <img src="https://img.shields.io/badge/-Agents-1F2937?style=flat-square" alt="Agents"> <img src="https://img.shields.io/badge/-Environments-1F2937?style=flat-square" alt="Environments">
  157. </p>
  158. Support custom pricing for custom and fine-tuned models.
  159. ---
  160. ## 🧪 Agent evals: LLM-as-a-judge, programmatic evals & CI gates
  161. Automatically evaluate LLM and agent outputs using LLM-as-a-Judge evaluations.
  162. Built-in evaluation types include:
  163. <p>
  164. <img src="https://img.shields.io/badge/-Hallucination-1F2937?style=flat-square" alt="Hallucination"> <img src="https://img.shields.io/badge/-Bias-1F2937?style=flat-square" alt="Bias"> <img src="https://img.shields.io/badge/-Toxicity-1F2937?style=flat-square" alt="Toxicity"> <img src="https://img.shields.io/badge/-Safety-1F2937?style=flat-square" alt="Safety"> <img src="https://img.shields.io/badge/-Instruction_following-1F2937?style=flat-square" alt="Instruction following"> <img src="https://img.shields.io/badge/-Completeness-1F2937?style=flat-square" alt="Completeness"> <img src="https://img.shields.io/badge/-Conciseness-1F2937?style=flat-square" alt="Conciseness"> <img src="https://img.shields.io/badge/-Sensitivity-1F2937?style=flat-square" alt="Sensitivity"> <img src="https://img.shields.io/badge/-Relevance-1F2937?style=flat-square" alt="Relevance"> <img src="https://img.shields.io/badge/-Coherence-1F2937?style=flat-square" alt="Coherence"> <img src="https://img.shields.io/badge/-Faithfulness-1F2937?style=flat-square" alt="Faithfulness">
  165. </p>
  166. Use evaluations to move from:
  167. **"The agent produced an answer."**
  168. to:
  169. **"The agent produced a good answer."**
  170. ---
  171. ## 🐛 Debug production AI agents
  172. Find the requests that matter.
  173. Investigate:
  174. <p>
  175. <img src="https://img.shields.io/badge/-Failed_LLM_calls-1F2937?style=flat-square" alt="Failed LLM calls"> <img src="https://img.shields.io/badge/-Exceptions-1F2937?style=flat-square" alt="Exceptions"> <img src="https://img.shields.io/badge/-Latency_spikes-1F2937?style=flat-square" alt="Latency spikes"> <img src="https://img.shields.io/badge/-Unexpected_costs-1F2937?style=flat-square" alt="Unexpected costs"> <img src="https://img.shields.io/badge/-Bad_evaluations-1F2937?style=flat-square" alt="Bad evaluations"> <img src="https://img.shields.io/badge/-Problematic_prompts-1F2937?style=flat-square" alt="Problematic prompts"> <img src="https://img.shields.io/badge/-Agent_tool_failures-1F2937?style=flat-square" alt="Agent/tool failures">
  176. </p>
  177. Go from:
  178. > `Something went wrong.`
  179. to a fully traced root cause:
  180. ```mermaid
  181. flowchart TD
  182. A[Agent] --> P[Prompt] --> L1[LLM] --> T[Tool call] --> R[Retrieval] --> L2[LLM] --> E([Error])
  183. style E fill:#DC2626,stroke:#7F1D1D,color:#fff
  184. style A fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  185. ```
  186. ---
  187. # 📸 See OpenLIT in action
  188. <table>
  189. <tr>
  190. <td width="33%" align="center">
  191. <img src="docs/images/docs-ai-observability-trace.png" alt="Traces view with full agent conversation, spans, and cost breakdown" width="100%"><br/>
  192. <sub><b>Traces</b> — full agent conversation, spans &amp; cost</sub>
  193. </td>
  194. <td width="33%" align="center">
  195. <img src="docs/images/auto-evals-dashboard.png" alt="Automated evaluation dashboard with hallucination, bias, and toxicity metrics" width="100%"><br/>
  196. <sub><b>Evaluations</b> — hallucination, bias &amp; toxicity checks</sub>
  197. </td>
  198. <td width="33%" align="center">
  199. <img src="docs/images/organisation/connectors-list.png" alt="Connectors catalog with ClickHouse, Grafana Tempo, Loki, Prometheus, and Jaeger" width="100%"><br/>
  200. <sub><b>Connectors</b> — plug in ClickHouse, Tempo, Loki, Prometheus &amp; more</sub>
  201. </td>
  202. </tr>
  203. <tr>
  204. <td width="33%" align="center">
  205. <img src="docs/images/docs-prompt-hub-details.png" alt="Prompt Hub prompt detail page with versions and linked rules" width="100%"><br/>
  206. <sub><b>Prompt Hub</b> — versioned, centrally managed prompts</sub>
  207. </td>
  208. <td width="33%" align="center">
  209. <img src="docs/images/rule-engine-conditions.png" alt="Rule Engine rule detail page with condition groups and live rule preview" width="100%"><br/>
  210. <sub><b>Rule Engine</b> — conditional rules on trace attributes</sub>
  211. </td>
  212. <td width="33%" align="center">
  213. <img src="docs/images/docs-ui-banner.jpeg" alt="OpenLIT dashboard preview with cost, latency, and usage charts" width="100%"><br/>
  214. <sub><b>Dashboards</b> — cost, latency &amp; usage charts at a glance</sub>
  215. </td>
  216. </tr>
  217. </table>
  218. ---
  219. # 🧠 Prompt management (Prompt Hub) & context
  220. Use **Prompt Hub** to:
  221. <p>
  222. <img src="https://img.shields.io/badge/-Create_prompts-1F2937?style=flat-square" alt="Create prompts"> <img src="https://img.shields.io/badge/-Version_prompts-1F2937?style=flat-square" alt="Version prompts"> <img src="https://img.shields.io/badge/-Retrieve_from_apps-1F2937?style=flat-square" alt="Retrieve from apps"> <img src="https://img.shields.io/badge/-Compare_versions-1F2937?style=flat-square" alt="Compare versions"> <img src="https://img.shields.io/badge/-Consistent_across_environments-1F2937?style=flat-square" alt="Consistent across environments">
  223. </p>
  224. Example:
  225. ```python
  226. prompt = openlit.prompts.get(
  227. "customer-support"
  228. )
  229. ```
  230. Keep prompt management separate from application code while maintaining version control and observability.
  231. Use **Context** to store reusable RAG content once and have the Rule Engine return the right piece at runtime. See the [Context docs](https://docs.openlit.io/latest/openlit/prompts-experiments/context/overview).
  232. ---
  233. # ⚙️ Rule Engine
  234. Define runtime rules based on trace attributes.
  235. Use rules to dynamically control:
  236. <p>
  237. <img src="https://img.shields.io/badge/-Prompts-1F2937?style=flat-square" alt="Prompts"> <img src="https://img.shields.io/badge/-Evaluations-1F2937?style=flat-square" alt="Evaluations"> <img src="https://img.shields.io/badge/-Contexts-1F2937?style=flat-square" alt="Contexts"> <img src="https://img.shields.io/badge/-Runtime_behavior-1F2937?style=flat-square" alt="Runtime behavior">
  238. </p>
  239. Example:
  240. ```text
  241. IF
  242. environment = production
  243. AND
  244. model = expensive-model
  245. THEN
  246. run cost evaluation
  247. + retrieve production prompt
  248. ```
  249. ---
  250. # 🛡️ Guardrails: prompt injection & sensitive-topic detection
  251. Use OpenLIT SDK guardrails to detect and block risky prompts at runtime:
  252. - Prompt injection / jailbreak attempts
  253. - Sensitive topics
  254. - Topic restriction
  255. Guardrail checks are traced with OpenTelemetry so you can see when and why a request was blocked. See the [guardrails docs](https://docs.openlit.io/latest/sdk/features/guardrails) and [quickstart](https://docs.openlit.io/latest/openlit/quickstart-guard).
  256. ---
  257. # 🔐 Vault: LLM API key management
  258. Store LLM API keys and other secrets in **Vault**, then retrieve them at runtime via the SDK or API — without hard-coding credentials in application code. See the [Vault docs](https://docs.openlit.io/latest/openlit/developer-resources/vault/overview).
  259. ---
  260. # 🆚 OpenGround: compare models side by side
  261. Run the same prompt across multiple providers in **OpenGround** and compare cost, latency, and output quality before you ship. See the [OpenGround docs](https://docs.openlit.io/latest/openlit/prompts-experiments/openground/overview).
  262. ---
  263. # 🖥️ GPU monitoring for LLM inference
  264. Monitor NVIDIA, AMD, and Intel GPUs used for LLM inference with the OpenTelemetry GPU collector: utilization, memory, power, and temperature, correlated with your traces. See the [GPU collector docs](https://docs.openlit.io/latest/gpu-collector/overview).
  265. ---
  266. # 🔌 OpenTelemetry-native (GenAI semantic conventions)
  267. OpenLIT is built around **OpenTelemetry**, rather than creating a proprietary telemetry format. Traces and metrics follow OpenTelemetry GenAI semantic conventions (`gen_ai.*`).
  268. Your telemetry can flow through the OpenTelemetry ecosystem:
  269. ```mermaid
  270. flowchart TD
  271. A["AI App / AI Agent"] -->|OTLP| R[OpenLIT OTLP receiver]
  272. A -->|optional sidecar| C[Your OpenTelemetry Collector]
  273. C --> R
  274. C --> O["Other OTel backends<br/>(Datadog, Grafana, Honeycomb, ...)"]
  275. R --> B[ClickHouse]
  276. B --> D[OpenLIT Dashboard]
  277. style A fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  278. style R fill:#F97316,stroke:#7C2D12,color:#fff
  279. style C fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  280. style B fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  281. style D fill:#F97316,stroke:#7C2D12,color:#fff
  282. ```
  283. OpenLIT listens for OTLP on `:4317` (gRPC) and `:4318` (HTTP). A bundled Collector is not required. You can still put your own Collector in front for fan-out or processing.
  284. This means you can integrate OpenLIT into an existing OpenTelemetry architecture instead of replacing it.
  285. ---
  286. # 🧩 70+ integrations
  287. OpenLIT auto-instruments a growing ecosystem of AI providers, frameworks, vector databases, and GPU infrastructure with a single line of code. Click any badge to view its integration guide.
  288. **LLM Providers**
  289. <a href="https://docs.openlit.io/latest/sdk/integrations/openai"><img src="https://img.shields.io/badge/OpenAI-1F2937?style=flat-square&logo=openai&logoColor=white" alt="OpenAI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/ollama"><img src="https://img.shields.io/badge/Ollama-1F2937?style=flat-square&logo=ollama&logoColor=white" alt="Ollama"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/anthropic"><img src="https://img.shields.io/badge/Anthropic-1F2937?style=flat-square&logo=anthropic&logoColor=white" alt="Anthropic"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/deepseek"><img src="https://img.shields.io/badge/DeepSeek-1F2937?style=flat-square&logo=deepseek&logoColor=white" alt="DeepSeek"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/gpt4all"><img src="https://img.shields.io/badge/GPT4All-1F2937?style=flat-square&logo=gpt4all&logoColor=white" alt="GPT4All"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/cohere"><img src="https://img.shields.io/badge/Cohere-1F2937?style=flat-square&logo=cohere&logoColor=white" alt="Cohere"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/mistral"><img src="https://img.shields.io/badge/Mistral-1F2937?style=flat-square&logo=mistralai&logoColor=white" alt="Mistral"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/github-models"><img src="https://img.shields.io/badge/GitHub_Models-1F2937?style=flat-square&logo=github&logoColor=white" alt="GitHub Models"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/vllm"><img src="https://img.shields.io/badge/vLLM-1F2937?style=flat-square&logo=vllm&logoColor=white" alt="vLLM"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/azure-openai"><img src="https://img.shields.io/badge/Azure_OpenAI-1F2937?style=flat-square&logo=azureopenai&logoColor=white" alt="Azure OpenAI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/azure-ai-inference"><img src="https://img.shields.io/badge/Azure_AI_Inference-1F2937?style=flat-square&logo=azureaiinference&logoColor=white" alt="Azure AI Inference"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/huggingface"><img src="https://img.shields.io/badge/HuggingFace-1F2937?style=flat-square&logo=huggingface&logoColor=white" alt="HuggingFace"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/bedrock"><img src="https://img.shields.io/badge/Amazon_Bedrock-1F2937?style=flat-square&logo=amazonbedrock&logoColor=white" alt="Amazon Bedrock"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/vertexai"><img src="https://img.shields.io/badge/Vertex_AI-1F2937?style=flat-square&logo=googlecloud&logoColor=white" alt="Vertex AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/google-ai-studio"><img src="https://img.shields.io/badge/Google_AI_Studio-1F2937?style=flat-square&logo=googlegemini&logoColor=white" alt="Google AI Studio"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/groq"><img src="https://img.shields.io/badge/Groq-1F2937?style=flat-square&logo=groq&logoColor=white" alt="Groq"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/nvidia-nim"><img src="https://img.shields.io/badge/NVIDIA_NIM-1F2937?style=flat-square&logo=nvidianim&logoColor=white" alt="NVIDIA NIM"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/xai"><img src="https://img.shields.io/badge/xAI-1F2937?style=flat-square&logo=xai&logoColor=white" alt="xAI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/elevenlabs"><img src="https://img.shields.io/badge/ElevenLabs-1F2937?style=flat-square&logo=elevenlabs&logoColor=white" alt="ElevenLabs"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/ai21"><img src="https://img.shields.io/badge/AI21-1F2937?style=flat-square&logo=ai21&logoColor=white" alt="AI21"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/together"><img src="https://img.shields.io/badge/Together_AI-1F2937?style=flat-square&logo=togetherai&logoColor=white" alt="Together AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/typesafe"><img src="https://img.shields.io/badge/TypeSafe_Jev-1F2937?style=flat-square&logoColor=white" alt="TypeSafe Jev"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/digitalocean"><img src="https://img.shields.io/badge/DigitalOcean_(pydo)-1F2937?style=flat-square&logo=digitalocean&logoColor=white" alt="DigitalOcean (pydo)"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/digitalocean-gradient"><img src="https://img.shields.io/badge/DigitalOcean_Gradient-1F2937?style=flat-square&logo=digitalocean&logoColor=white" alt="DigitalOcean Gradient"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/assemblyai"><img src="https://img.shields.io/badge/Assembly_AI-1F2937?style=flat-square&logo=assemblyai&logoColor=white" alt="Assembly AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/featherless"><img src="https://img.shields.io/badge/Featherless-1F2937?style=flat-square&logo=featherless&logoColor=white" alt="Featherless"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/reka"><img src="https://img.shields.io/badge/Reka_AI-1F2937?style=flat-square&logo=rekaai&logoColor=white" alt="Reka AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/krutrim"><img src="https://img.shields.io/badge/OLA_Krutrim-1F2937?style=flat-square&logo=olakrutrim&logoColor=white" alt="OLA Krutrim"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/titan-ml"><img src="https://img.shields.io/badge/Titan_ML-1F2937?style=flat-square&logo=titanml&logoColor=white" alt="Titan ML"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/sarvam"><img src="https://img.shields.io/badge/Sarvam_AI-1F2937?style=flat-square&logo=sarvamai&logoColor=white" alt="Sarvam AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/premai"><img src="https://img.shields.io/badge/Prem_AI-1F2937?style=flat-square&logo=premai&logoColor=white" alt="Prem AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/replicate"><img src="https://img.shields.io/badge/Replicate-1F2937?style=flat-square&logo=replicate&logoColor=white" alt="Replicate"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/oci"><img src="https://img.shields.io/badge/OCI_GenAI-1F2937?style=flat-square&logo=ocigenai&logoColor=white" alt="OCI GenAI"></a>
  290. **Vector & Data Stores**
  291. <a href="https://docs.openlit.io/latest/sdk/integrations/chromadb"><img src="https://img.shields.io/badge/ChromaDB-1F2937?style=flat-square&logo=chromadb&logoColor=white" alt="ChromaDB"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/pinecone"><img src="https://img.shields.io/badge/Pinecone-1F2937?style=flat-square&logo=pinecone&logoColor=white" alt="Pinecone"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/qdrant"><img src="https://img.shields.io/badge/Qdrant-1F2937?style=flat-square&logo=qdrant&logoColor=white" alt="Qdrant"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/milvus"><img src="https://img.shields.io/badge/Milvus-1F2937?style=flat-square&logo=milvus&logoColor=white" alt="Milvus"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/astradb"><img src="https://img.shields.io/badge/AstraDB-1F2937?style=flat-square&logo=datastax&logoColor=white" alt="AstraDB"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/psycopg"><img src="https://img.shields.io/badge/PostgreSQL_(psycopg3)-1F2937?style=flat-square&logo=postgresql&logoColor=white" alt="PostgreSQL (psycopg3)"></a>
  292. **AI Frameworks & Agents**
  293. <a href="https://docs.openlit.io/latest/sdk/integrations/langchain"><img src="https://img.shields.io/badge/LangChain-1F2937?style=flat-square&logo=langchain&logoColor=white" alt="LangChain"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/openai-agents"><img src="https://img.shields.io/badge/OpenAI_Agents-1F2937?style=flat-square&logo=openaiagents&logoColor=white" alt="OpenAI Agents"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/litellm"><img src="https://img.shields.io/badge/LiteLLM-1F2937?style=flat-square&logo=litellm&logoColor=white" alt="LiteLLM"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/crewai"><img src="https://img.shields.io/badge/CrewAI-1F2937?style=flat-square&logo=crewai&logoColor=white" alt="CrewAI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/llama-index"><img src="https://img.shields.io/badge/LlamaIndex-1F2937?style=flat-square&logo=llamaindex&logoColor=white" alt="LlamaIndex"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/browser-use"><img src="https://img.shields.io/badge/Browser_Use-1F2937?style=flat-square&logo=browseruse&logoColor=white" alt="Browser Use"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/pydantic"><img src="https://img.shields.io/badge/Pydantic_AI-1F2937?style=flat-square&logo=pydantic&logoColor=white" alt="Pydantic AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/dspy"><img src="https://img.shields.io/badge/DSPy-1F2937?style=flat-square&logo=dspy&logoColor=white" alt="DSPy"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/ag2"><img src="https://img.shields.io/badge/AutoGen_(AG2)-1F2937?style=flat-square&logo=autogen&logoColor=white" alt="AutoGen (AG2)"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/haystack"><img src="https://img.shields.io/badge/Haystack-1F2937?style=flat-square&logo=haystack&logoColor=white" alt="Haystack"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/mem0"><img src="https://img.shields.io/badge/mem0-1F2937?style=flat-square&logo=mem0&logoColor=white" alt="mem0"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/guardrails"><img src="https://img.shields.io/badge/Guardrails_AI-1F2937?style=flat-square&logo=guardrailsai&logoColor=white" alt="Guardrails AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/phidata"><img src="https://img.shields.io/badge/Phidata-1F2937?style=flat-square&logo=phidata&logoColor=white" alt="Phidata"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/multion"><img src="https://img.shields.io/badge/MultiOn-1F2937?style=flat-square&logo=multion&logoColor=white" alt="MultiOn"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/julep-ai"><img src="https://img.shields.io/badge/Julep_AI-1F2937?style=flat-square&logo=julepai&logoColor=white" alt="Julep AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/letta"><img src="https://img.shields.io/badge/Letta-1F2937?style=flat-square&logo=letta&logoColor=white" alt="Letta"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/crawl4ai"><img src="https://img.shields.io/badge/Crawl4AI-1F2937?style=flat-square&logo=crawl4ai&logoColor=white" alt="Crawl4AI"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/firecrawl"><img src="https://img.shields.io/badge/FireCrawl-1F2937?style=flat-square&logo=firecrawl&logoColor=white" alt="FireCrawl"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/dynamiq"><img src="https://img.shields.io/badge/Dynamiq-1F2937?style=flat-square&logo=dynamiq&logoColor=white" alt="Dynamiq"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/controlflow"><img src="https://img.shields.io/badge/ControlFlow-1F2937?style=flat-square&logo=controlflow&logoColor=white" alt="ControlFlow"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/swarmzero"><img src="https://img.shields.io/badge/SwarmZero-1F2937?style=flat-square&logo=swarmzero&logoColor=white" alt="SwarmZero"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/langgraph"><img src="https://img.shields.io/badge/LangGraph-1F2937?style=flat-square&logo=langgraph&logoColor=white" alt="LangGraph"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/strands"><img src="https://img.shields.io/badge/Strands_Agents-1F2937?style=flat-square&logo=strandsagents&logoColor=white" alt="Strands Agents"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/google-adk"><img src="https://img.shields.io/badge/Google_ADK-1F2937?style=flat-square&logo=google&logoColor=white" alt="Google ADK"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/claude-agent-sdk"><img src="https://img.shields.io/badge/Claude_Agent_SDK-1F2937?style=flat-square&logo=anthropic&logoColor=white" alt="Claude Agent SDK"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/vercel-ai"><img src="https://img.shields.io/badge/Vercel_AI_SDK-1F2937?style=flat-square&logo=vercel&logoColor=white" alt="Vercel AI SDK"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/agent-framework"><img src="https://img.shields.io/badge/Agent_Framework-1F2937?style=flat-square&logo=agentframework&logoColor=white" alt="Agent Framework"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/smolagents"><img src="https://img.shields.io/badge/smolagents-1F2937?style=flat-square&logo=huggingface&logoColor=white" alt="smolagents"></a>
  294. **Governance & Protocols**
  295. <a href="https://docs.openlit.io/latest/sdk/integrations/agent-governance-toolkit"><img src="https://img.shields.io/badge/Agent_Governance_Toolkit-1F2937?style=flat-square&logo=agentgovernancetoolkit&logoColor=white" alt="Agent Governance Toolkit"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/mcp"><img src="https://img.shields.io/badge/MCP-1F2937?style=flat-square&logo=mcp&logoColor=white" alt="MCP"></a>
  296. **GPU Monitoring**
  297. <a href="https://docs.openlit.io/latest/sdk/integrations/nvidia-gpu"><img src="https://img.shields.io/badge/NVIDIA_GPUs-1F2937?style=flat-square&logo=nvidia&logoColor=white" alt="NVIDIA GPUs"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/amd-gpu"><img src="https://img.shields.io/badge/AMD_GPUs-1F2937?style=flat-square&logo=amd&logoColor=white" alt="AMD GPUs"></a> <a href="https://docs.openlit.io/latest/sdk/integrations/intel-gpu"><img src="https://img.shields.io/badge/Intel_GPUs-1F2937?style=flat-square&logo=intel&logoColor=white" alt="Intel GPUs"></a>
  298. See the complete integration list in the [documentation](https://docs.openlit.io/).
  299. ---
  300. # 🛠️ SDKs
  301. OpenLIT provides OpenTelemetry-native SDKs for:
  302. ### Python
  303. ```bash
  304. pip install openlit
  305. ```
  306. [Python SDK →](https://github.com/openlit/openlit/tree/main/sdk/python)
  307. ### TypeScript
  308. ```bash
  309. npm install openlit
  310. ```
  311. [TypeScript SDK →](https://github.com/openlit/openlit/tree/main/sdk/typescript)
  312. ### Go
  313. [Go SDK →](https://github.com/openlit/openlit/tree/main/sdk/go)
  314. ---
  315. # 🏗️ Architecture
  316. OpenLIT is designed to run in your infrastructure.
  317. A typical deployment looks like:
  318. ```mermaid
  319. flowchart TD
  320. subgraph App["Your application"]
  321. direction LR
  322. Agent --> LLM --> Tools --> RAG --> DB
  323. end
  324. App -->|OTLP| Receiver[OpenLIT OTLP receiver]
  325. Receiver --> CH[(ClickHouse)]
  326. CH --> Dash[OpenLIT Dashboard]
  327. style App fill:#1F2937,stroke:#F97316,stroke-width:2px,color:#fff
  328. style Receiver fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  329. style CH fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  330. style Dash fill:#F97316,stroke:#7C2D12,color:#fff
  331. ```
  332. ---
  333. # 🔐 Self-hosted by default
  334. Run OpenLIT inside your own infrastructure using Docker or Kubernetes.
  335. Your telemetry stays under your control.
  336. ```bash
  337. docker compose up -d
  338. ```
  339. For Kubernetes, see the [installation documentation](https://docs.openlit.io/latest/openlit/installation#kubernetes).
  340. ---
  341. # 🚀 From trace to optimization
  342. Observability is only the beginning.
  343. OpenLIT is designed around a continuous AI engineering loop:
  344. ```mermaid
  345. flowchart LR
  346. T[Trace] --> E[Evaluate] --> A[Analyze] --> O[Optimize] --> M[Manage] -.-> T
  347. style T fill:#F97316,stroke:#7C2D12,color:#fff
  348. style E fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  349. style A fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  350. style O fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  351. style M fill:#111827,stroke:#F97316,stroke-width:2px,color:#fff
  352. ```
  353. The goal is simple:
  354. > **Make AI systems observable, measurable, debuggable, and continuously improvable.**
  355. ---
  356. ## ❓ FAQ
  357. **What is OpenLIT?**
  358. OpenLIT is an open-source agent harness engineering platform. It provides OpenTelemetry-native tracing, evaluations, guardrails, prompt management, and cost and GPU monitoring for AI agents and coding agents, and it is free to self-host under Apache 2.0.
  359. **What is an agent harness?**
  360. An agent harness is everything in an AI agent except the model: the tools, context, prompts, memory, hooks, guardrails, and feedback loops that turn a model into a working agent. Claude Code, Codex, and frameworks like LangGraph or CrewAI are harnesses.
  361. **What is agent harness engineering?**
  362. Agent harness engineering is the discipline of designing, measuring, and improving the harness around a model so agents are reliable in production. Teams run agents on real tasks, observe failures in traces, evaluate them, fix the harness (prompts, tools, rules, guardrails), and verify the fix with regression evals.
  363. **Is OpenLIT an agent framework or a harness runtime?**
  364. No. OpenLIT doesn't run your agent loop. It instruments and improves whatever harness you already use, through OpenTelemetry, so you can switch models, frameworks, or harnesses without losing your traces, evals, or prompts.
  365. **How is OpenLIT different from Langfuse, LangSmith, or Arize Phoenix?**
  366. OpenLIT is Apache-2.0 and OpenTelemetry-native end to end, adds GPU monitoring, guardrails, Vault, and coding-agent observability in the same self-hosted platform, and exports to any OTLP backend. See https://openlit.io/compare for feature-by-feature comparisons.
  367. **Does OpenLIT work with Claude Code, Codex, and Cursor?**
  368. Yes. The `openlit` CLI ingests each coding agent's hook events and maps them to OpenTelemetry `gen_ai.*` conventions, with no SDK and no code changes.
  369. **Can I run agent evals in CI?**
  370. Yes. Run LLM-as-a-judge and programmatic evals online on production traces, or offline through the SDK as CI/CD gates.
  371. **Is OpenLIT free?**
  372. Yes. Self-hosted OpenLIT is free under Apache 2.0, with no license key and no per-trace fee. Organisation-wide governance features such as custom roles, audit logs, and alerts are part of [OpenLIT Enterprise](https://docs.openlit.io/latest/openlit/enterprise).
  373. **Does OpenLIT add latency?**
  374. No proxy is required. SDKs instrument in-process and export telemetry asynchronously over OTLP.
  375. **Can I send data to Grafana, Datadog, or my existing OpenTelemetry Collector?**
  376. Yes. OpenLIT emits standard OTLP traces and metrics, so they can go to OpenLIT, Grafana, Datadog, New Relic, SigNoz, or any OTLP backend.
  377. ---
  378. ## 🆚 How OpenLIT compares
  379. Feature-by-feature comparisons on [openlit.io/compare](https://openlit.io/compare):
  380. - [OpenLIT vs Langfuse](https://openlit.io/compare/openlit-vs-langfuse)
  381. - [OpenLIT vs LangSmith](https://openlit.io/compare/openlit-vs-langsmith)
  382. - [OpenLIT vs Arize Phoenix](https://openlit.io/compare/openlit-vs-arize-phoenix)
  383. - [OpenLIT vs Braintrust](https://openlit.io/compare/openlit-vs-braintrust)
  384. - [OpenLIT vs Helicone](https://openlit.io/compare/openlit-vs-helicone)
  385. - [OpenLIT vs Datadog](https://openlit.io/compare/openlit-vs-datadog)
  386. - [OpenLIT vs Comet Opik](https://openlit.io/compare/openlit-vs-comet-opik)
  387. ---
  388. # 🌎 Community
  389. OpenLIT is open source and built with the AI engineering community.
  390. Join us:
  391. * [GitHub](https://github.com/openlit/openlit)
  392. * [Documentation](https://docs.openlit.io/)
  393. * [Slack](https://join.slack.com/t/openlit/shared_invite/zt-2etnfttwg-TjP_7BZXfYg84oAukY8QRQ)
  394. * [X / Twitter](https://twitter.com/openlit_io)
  395. If OpenLIT is useful to you, **please consider giving the repository a ⭐**.
  396. It helps other AI engineers discover the project.
  397. ---
  398. # 🤝 Contributing
  399. Contributions are welcome.
  400. You can contribute by:
  401. * fixing bugs
  402. * adding integrations
  403. * improving documentation
  404. * creating examples
  405. * improving SDKs
  406. * adding evaluations
  407. * building dashboards
  408. * reporting issues
  409. * sharing OpenLIT with other developers
  410. Check the repository's issues for opportunities to contribute. To run OpenLIT from source, see [Run OpenLIT locally](CONTRIBUTING.md#run-openlit-locally): one script, `scripts/dev-realtime.sh`, starts the UI, OTLP receiver, realtime engine, and NATS.
  411. ---
  412. # 📄 License
  413. OpenLIT is licensed under the Apache License 2.0.
  414. See [LICENSE](https://github.com/openlit/openlit/blob/main/LICENSE) for details.
  415. ---
  416. # 🙇 Acknowledgments
  417. ## Sponsors
  418. ### Silver
  419. <p>
  420. <a href="https://fluxionai.world/register?source=github&campaign=github-openlit&promo=OPENLIT" target="_blank">
  421. <img src="docs/images/fluxion-ai-logo.png" alt="Fluxion AI" height="72">
  422. </a>
  423. </p>
  424. ### Bronze
  425. <p>
  426. <a href="https://www.testmuai.com/?utm_medium=sponsor&utm_source=openlit" target="_blank">
  427. <img src="docs/images/testmu-logo.png" alt="TestMu AI" height="80">
  428. </a>
  429. </p>
  430. ## Deployment partners
  431. <p>
  432. <a href="https://www.digitalocean.com/">
  433. <img src="https://opensource.nyc3.cdn.digitaloceanspaces.com/attribution/assets/SVG/DO_Logo_horizontal_blue.svg" alt="DigitalOcean" height="80" width="200">
  434. </a>
  435. &nbsp;&nbsp;&nbsp;&nbsp;
  436. <a href="https://www.hostg.xyz/aff_c?offer_id=815&aff_id=243668&url_id=6792">
  437. <img src="https://assets.hostinger.com/vps/deploy.svg" alt="Deploy on Hostinger">
  438. </a>
  439. </p>
  440. ---
  441. # 💻 Contributors
  442. <a href="https://github.com/openlit/openlit/graphs/contributors">
  443. <img src="https://contrib.rocks/image?repo=openlit/openlit&max=500&columns=20" alt="OpenLIT contributors" width="100%">
  444. </a>
  445. Names for the same contributors are on [openlit.io/about-us](https://openlit.io/about-us).
  446. ---
  447. <div align="center">
  448. ### Build agent harnesses you can actually measure and improve.
  449. **⭐ Star OpenLIT on GitHub**
  450. </div>

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.