Build with Pendra

Pendra is a UK-sovereign AI inference platform. The API speaks OpenAI and Anthropic shapes, the SDKs are drop-in replacements, and the worker runs on your own hardware when you want full data residency. Pick a section to dive in.

Quickstart

Install an SDK, grab an API key, and make your first request — under five minutes.

API reference

Chat, embeddings, images, audio, and OpenAI/Anthropic-compatible surfaces. Same shape as the OpenAI and Anthropic APIs you already use.

Integrations

Drop-in setup for Claude Code, OpenAI Codex, and OpenClaw against Pendra.

Self-hosted Worker

Run inference on your own GPUs. Supported OSes, system requirements, install paths.

Backend capabilities

Which inference backends Pendra connects to, what each supports, and where curated catalogue installs are available.

SDKs

Python, Node, Go, Rust, and .NET clients. OpenAI-compatible interfaces.

What's where

Common paths through these docs:

Building against the API? Start with the Quickstart, then the API reference.
Pointing a coding agent at Pendra? Jump straight to Claude Code, Codex, or OpenClaw.
Hosting inference yourself? See the Worker overview and the backend capability matrix.
Looking for a specific SDK? Python, Node, Go, Rust, .NET.