OpenAI-compatible · distributed

Konstellation

Run inference across a constellation of community GPUs — one OpenAI-compatible API.

Use the API Run a node & earn Console

For developers

Point the OpenAI SDK at Konstellation. Streaming, embeddings, usage — same shapes.

For node owners

Donate your Mac/PC GPU, get verified, and earn 70% of the revenue for traffic you serve.

Quickstart — consumers

1. Get an API key (dev key: sk-konst-devkey). 2. Set the base URL. 3. Call it.

curl $HUB/v1/chat/completions \
  -H "Authorization: Bearer sk-konst-devkey" \
  -H "Content-Type: application/json" \
  -d '{"model":"echo/demo","messages":[{"role":"user","content":"Hello!"}]}'

from openai import OpenAI
client = OpenAI(base_url="$HUB/v1", api_key="sk-konst-devkey")
client.chat.completions.create(
    model="echo/demo",
    messages=[{"role":"user","content":"Hello!"}],
)

Streaming with "stream": true; usage is metered Hub-side and shown in the console. Full reference: docs/api.md.

Quickstart — node owners

# 1. Download the agent for your OS (see Releases), then:
konst-agent enroll <enrollment-token>     # token from the console / operator
KONST_HUB_URL=wss://$HUB konst-agent run   # serves jobs, sends heartbeats

The agent tunnels out (no inbound ports), advertises your local models (Ollama), and is verified before it receives traffic. Earnings accrue per token served and pay out monthly (min $25). See payout details.

Pricing

Live rates from the active plan (per 1M tokens):

Model	Input	Output
Loading… (or see the console)