Developers/ Integrations
DEVELOPERS · INTEGRATIONS

Bring your stack. Point it at ARK.

ARK exposes an OpenAI v1 and Anthropic-compatible surface, so the agents and chat frameworks you already use work without code changes. Four best-in-class open-source stacks are verified end-to-end — pick one, change the base URL, ship.

Start free → Read the API docs
OpenAI v1
Compatible
Drop-in API
Anthropic
Messages API
Compatible
EU-hosted
Zero retention
by default
−98.9%
Tokens/call in
stateful mode
THE FOUR STACKS

Four production-grade open-source stacks. One sovereign runtime.

Each card links to a setup guide that gets you from git clone to a live agent talking to ARK in under five minutes. No wrappers, no lock-in, no data leaving the region.

Personal · Multi-channel

OpenClaw

Config-first personal AI agent that runs on any OS and speaks in the channels you already use — WhatsApp, Slack, Telegram, Discord, iMessage, and 20+ more. Write a SOUL.md, run one command, it's live.

Multi-channel Config-first Self-hosted
github.com/openclaw/openclaw
Enterprise · Secure

NVIDIA NemoClaw

NVIDIA's enterprise wrapper around OpenClaw. Adds OpenShell sandboxing, policy-based guardrails, and lifecycle management — so autonomous agents run safely inside regulated environments.

On-prem ready Sandboxed NVIDIA-backed
github.com/NVIDIA/NemoClaw
Knowledge · RAG

AnythingLLM

Private knowledge workspace with built-in RAG, document ingestion, citations, and multi-user workspaces. Point it at ARK and your documents never leave the region.

RAG Multi-user Self-hosted
github.com/Mintplex-Labs/anything-llm
Autonomous · Self-improving

Hermes Agent

A self-improving autonomous agent with long-running memory, tool use, and reflection loops. Built for operators who want an assistant that gets better the longer it runs.

Memory Tool use Long-running
github.com/hermes-agent/hermes
ROLL YOUR OWN

Already have a stack? Point it at ARK.

ARK exposes OpenAI v1 and Anthropic Messages-compatible endpoints. If your code already calls one of those, you're one base-URL change away from sovereign inference.

API REFERENCE

Documentation

Auth, endpoints, streaming, tool calling, stateful sessions, embeddings, audio, and image generation. Every feature, every parameter, every example.

/developers/docs/
BENCHMARKS

Research & Benchmarks

Performance data, test suite methodology, and head-to-head numbers against hyperscaler endpoints — so you know what to expect before you migrate.

/developers/research/