Skip to main content
Are you a large language model? This page is available as raw markdown at /index.md. The full docset is at /llms-full.md and the index is at /llms.md.

Quickstart

C3 runs your script on a cloud GPU or CPU and returns the results. You describe the job in a small .c3 file; C3 finds the hardware, runs the script and collects the output. Use it from a coding agent through the MCP server, or from your terminal with the CLI.

From a coding agent​

C3 hosts an MCP server at https://api.cthree.cloud/mcp. An agent connected to it can list hardware, deploy jobs, follow logs and collect results. Nothing is installed where the agent runs.

1. Get an API key from dashboard settings, or run c3 apikey create my-agent if you have the CLI.

2. Connect your host. For Claude Code:

claude mcp add --transport http c3 https://api.cthree.cloud/mcp \
  --header "Authorization: Bearer c3_key_..."

Cursor, Codex, VS Code, ChatGPT and claude.ai setup is on the MCP server page.

3. Ask for work. For example: "List the hardware C3 offers, deploy run.sh from this folder on an l40, follow it, and fetch the results."

From the terminal​

1. Install and sign in.

curl -fsSL https://cthree.cloud/install.sh | sh
c3 login

c3 login opens your browser. Verify your email when prompted; jobs need a verified address.

2. Run the example.

git clone https://github.com/c3-research/c3-examples.git
cd c3-examples/jax-matmul
c3 deploy -f

The .c3 file in that folder is the whole job description:

project: jax-matmul-example
script: run.sh          # a bash script with your commands
hardware: l40           # l40, a100 or h100
time: "00:10:00"        # maximum runtime

python:
  project: ./           # installs pyproject.toml dependencies with uv

output:
  - ./results           # folders collected as results

C3 provisions the machine, installs the Python dependencies, runs run.sh and uploads results/. The -f flag streams logs until the job finishes.

3. Check status and collect results.

c3 squeue              # jobs and their status
c3 pull job_abc123     # downloads ./job_abc123/results/

4. Run your own code. In your project folder, run c3 init, set script, hardware and time in the generated .c3, then c3 deploy.

Next​