Klu documentation
Use Klu to build AI applications with multiple model providers, run them through an API or SDK, inspect production Logs, and improve quality with Datasets and evaluations.
Last updated: August 18, 2026
Guides
Build your first app
Create a Workspace, connect a model provider, and build an Action in Studio.
Create an Action
Configure a prompt, model, variables, and output for reusable AI functionality.
Integrate an OpenAI app
Route an existing OpenAI-compatible integration through the Klu gateway.
Run evaluations
Create test cases and compare prompt and model performance with an evaluation.
Work with Datasets
Create, import, curate, and reuse examples for evaluation and fine-tuning.
Run Klu in production
Configure, deploy, monitor, and troubleshoot the Klu runtime services.
What you can do with Klu
Klu organizes your work into Workspaces, Apps, and Actions. You create and compare prompts in Studio, connect model providers and external services, add Context for retrieval, and publish an Action for use from your application.
Once an Action is in use, Klu records its generations as Logs. You can review feedback, collect examples in Datasets, run evaluations, view Insights, and fine-tune supported models.
Start with key concepts if you are new to the platform. If you already have a Workspace and API key, go directly to the API and SDK basics.
Model providers
You connect a model provider at the Workspace level, then choose an available provider and model in Studio. The current app includes these provider options:
| Standard provider options | Scale and custom provider options |
|---|---|
| OpenAI, Anthropic, AI21, Cohere | Amazon Bedrock, Azure OpenAI, Azure AI Endpoint |
| Google AI, Perplexity AI, Together AI | GCP Vertex, Cloudflare AI, Hugging Face Hub |
| Mistral, Groq, DeepSeek, Grok | Replicate, API Endpoint |
| Klu |
Provider availability, models, and features can vary by Workspace configuration and by the credentials you supply. Check the provider and model selectors in Connections and Studio for the options available to your Workspace.
Connections, Context, and Skills
Use Connections to authorize services that Klu can use for Context or Skills. The current connection catalog includes Asana, Airtable, GitHub, Google, Intercom, Jira, Notion, Slack, Teams, Zendesk, and Zoom.
Use Context to store and retrieve information that grounds an Action. Context sources can include uploaded files, configured databases, and API responses. The exact sources available depend on your Workspace configuration and deployment.
Use Skills to let an Action call a connected API or execute configured compute. Treat Context as information the model can retrieve and a Skill as an operation the model can invoke.
Choose your next step
Resources
API and SDK basics
Authenticate, install the Python or TypeScript SDK, and make your first request.
Actions API
Run deployed Actions, pass variables and metadata, and handle responses.
Context API
Create Context, add documents, and retrieve grounded information.
SDK exports
See the clients exposed by the current Python and TypeScript SDKs.
Apps and Data API
Create Apps and manage generation Data through the public API.
Evaluation and optimization API
Work with Datasets, Evals, Experiments, Feedback, and fine-tunes.
Files and Context sources API
Upload files and manage Documents and Context sources programmatically.
Models and Connections API
Manage models, provider Connections, and Workspaces through the API.
Sessions, Skills, and Workflows API
Use the public endpoints for Sessions, Skills, and Workflows.