Skip to main content
Use this map to find the part of Planck that matches the problem you are solving. If you have not sent a request yet, begin with the quickstart.

Models and inference

Call supported models through OpenAI-compatible or Anthropic-compatible APIs.

Chat & Responses

Generate text with familiar request and response formats.

Model capabilities

Find models by operation, modality, parameter, and provider endpoint.

Provider routing

Choose a provider or let Planck route across compatible endpoints.

Tool calling

Let models request application-defined functions.

Structured JSON

Constrain model output to a schema your application can validate.

Media and data

Work with embeddings, reranking, images, video, audio, and files.

Reliability and controls

Apply production behavior without rebuilding each provider integration.

Fallbacks

Understand eligible retries, provider failures, and fallback behavior.

Response caching

Reuse eligible gateway responses to reduce latency and cost.

Prompt caching

Reuse provider-side prompt context on supported endpoints.

Rate limits

Control request, cost, or property-based usage.

Prompt security

Detect prompt attacks inline and handle blocked requests.

Gateway headers

Configure supported request-level gateway behavior.

Observability

Connect model activity to the users, sessions, and application context that produced it.

Requests

Inspect inputs, outputs, cost, latency, errors, and provider details.

Properties

Add consistent metadata for filtering and analysis.

Sessions

Group related calls into conversations and agent workflows.

Scores & feedback

Attach automated evaluations and end-user feedback to requests.

Datasets

Curate reusable request sets for evaluation and export.

Webhooks

Deliver completed request events to your systems.

Prompt management

Version, assemble, and deploy prompts independently from application releases.

Prompt library

Create reusable, versioned prompt definitions.

Gateway deployment

Select and render managed prompts in gateway requests.

Prompt assembly

Understand templates, variables, messages, and runtime parameters.

Integrations

Use Planck as the request path, add it to an existing provider integration, or record non-model steps in a larger trace.

Choose an integration

Compare gateway, automatic logging, manual logging, and custom traces.

Frameworks

Connect Codex, LangChain, Semantic Kernel, Vercel AI SDK, and more.

Custom traces

Record tools, retrieval, and application-specific operations.

Keep exploring

Guides

Follow end-to-end application and prompting walkthroughs.

API reference

Inspect gateway and platform endpoints in detail.