BUILD / DOCUMENTATION

Your version.
Your docs.

Start building with vLLM Agentic API. Choose the documentation that matches the version you run.

Documentation version

MAIN BRANCH

Rust command-line reference

Latest / development · The agentic CLI: run harnesses, serve, and validate.

Latest / development · The agentic-api Python launcher: commands, options, and defaults.

Install the published package, connect a running model, and send a first Responses API request.

Explore the Responses, Messages, and other HTTP API endpoints.

Install from PyPI, check the package, and choose a remote or managed vLLM workflow.

Connect coding clients and verify their setup against an Agentic API gateway.

Compact Responses item history and continue with the resulting context window.

Configure Codex Desktop to use a local model through Agentic API.

Connect an SGLang inference endpoint and review its tested configuration.

Connect NVIDIA Dynamo and a vLLM worker to the gateway.

Build and configure the production gateway container.

Export gateway traces and metrics with OpenTelemetry.

Set up a contributor development environment and build the repository.

Find ways to participate and contribute to the project.

Explore the v1alpha response hydration and persistence design.

Configure the gateway and its dependencies on Kubernetes.

Run a local Kubernetes deployment with kind.

Set up an authenticated gateway deployment with Dex.

Prepare, validate, and publish project releases.