Get started in 60 seconds

Pick how you want to use cocompute. You can do both.

Host

Share your GPU

You have a GPU sitting idle. Install the host binary, join the pool, and get access to other people's GPUs in return.

Jump to steps →
Consumer

Use the pool

You want LLM inference but don't want to buy hardware. Sign up, get an API key, point your apps at our OpenAI-compatible endpoint.

Jump to steps →
Host

Share your GPU

Three steps. About 5 minutes if you already have Ollama installed.

1

Sign up

Create a free account so you can manage your hosts and pools.

2

Install the host binary

From your dashboard, click Add Host to get a one-line install command. Run it on any machine that runs Ollama:

curl -sSf https://cocompute.ai/install.sh | COCOMPUTE_URL=https://cocompute.ai bash -s -- --token YOUR_TOKEN

Works on Linux (systemd) and macOS (launchd). Runs as a background service. Anything Ollama supports works: NVIDIA, AMD, Apple Silicon, even CPU.

3

Add your host to a pool

Back in the dashboard, create a pool (or pick the global pool) and add your host. As soon as your host registers, it shows up online and is ready to serve inference.

Share with friends, your team, or the global pool. No tokens, no crypto.

Consumer

Use the pool

Three steps. About 2 minutes if you have a curl handy.

1

Sign up

Create a free account.

2

Create an API key

From your dashboard, find the pool you want to use and click New API Key. Copy the key. It looks like a long random string. You'll only see it once.

API keys are scoped to a pool. They only have access to the hosts in that pool.

3

Make your first call

Drop your key in and call the OpenAI-compatible endpoint:

curl https://cocompute.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gemma4:31b","messages":[{"role":"user","content":"hello"}]}'

Or list which models the pool has available:

curl https://cocompute.ai/v1/models -H "Authorization: Bearer YOUR_API_KEY"

Works with any client that speaks the OpenAI API spec (the official OpenAI SDK, openwebui, llama.cpp clients). Just change the base URL.

Stuck?

Open an issue on GitHub or read the source. cocompute is AGPL, every line is yours to inspect.

Sign upView sourceBack to home