Integration guide

OpenClaw

OpenClaw is a personal assistant that runs on your own machine and answers you in Discord, Slack, Telegram, WhatsApp and more. This guide adds legit2compute as an OpenClaw model provider and installs our media skill, so your assistant can render images, music, voice and video on the network.

Base URL (OpenAI-compatible)

https://api.legit2compute.com/nodes/llm/v1

API key

l2ck_… from /developers

Recommended model

qwen3.8:27b

Context window

131,072 tokens

Sign in and the snippets on this page fill in your own key.

1. Install OpenClaw

Follow the OpenClaw README. On Linux, macOS or WSL2:

shell

curl -fsSL https://openclaw.ai/install.sh | bash
openclaw onboard --install-daemon

Already running Node 24 or newer? npm does it too:

shell

npm install -g openclaw@latest --allow-scripts=openclaw

2. Store your key

Copy your l2ck_ key from legit2compute.com/developers into OpenClaw's global env file. The gateway reads it even when it runs as a background service, and the media skill picks it up from there too.

shell

echo 'L2C_API_KEY=l2ck_...' >> ~/.openclaw/.env

3. Add legit2compute as a provider

Add this to ~/.openclaw/openclaw.json (JSON5; the gateway reloads it when you save). The provider ID legit2compute is yours to choose; model references are provider/model.

~/.openclaw/openclaw.json

{
  models: {
    mode: "merge",
    providers: {
      legit2compute: {
        baseUrl: "https://api.legit2compute.com/nodes/llm/v1",
        apiKey: "${L2C_API_KEY}",
        api: "openai-completions",
        models: [
          { id: "qwen3.8:27b", name: "Qwen3.8 27B (vision + text)",
            input: ["text", "image"], contextWindow: 131072 },
          { id: "llama3.1", name: "Llama 3.1 8B",
            input: ["text"], contextWindow: 131072 },
          { id: "ornith-1.5:9b", name: "Ornith 1.5 9B (vision, 256K)",
            input: ["text", "image"], contextWindow: 262144 },
        ],
      },
    },
  },
  agents: {
    defaults: {
      model: { primary: "legit2compute/qwen3.8:27b" },
    },
  },
}

Or set the default from the command line, and confirm the provider loaded:

shell

openclaw models set legit2compute/qwen3.8:27b
openclaw models list

Then check one turn end to end. A brand-new assistant may first ask what to call it; that is OpenClaw's own first-run setup, and the reply proves the connection works.

shell

openclaw agent --local -m "Reply with exactly one word: hello"

Prefer editing in a browser? openclaw dashboard opens the Control UI, whose Config tab edits the same file.

4. Install the media skill

OpenClaw follows the Agent Skills standard and loads skills shared by all your agents from ~/.openclaw/skills:

shell

mkdir -p ~/.openclaw/skills
curl -fsSL -o /tmp/legit2cloud-skill.zip https://legit2compute.com/developers/skill.zip
unzip -o /tmp/legit2cloud-skill.zip -d ~/.openclaw/skills/

The skill needs python3 on the machine running the gateway and reads L2C_API_KEY, which you already put in ~/.openclaw/.env. openclaw skills list should show legit2cloud as ready. Then ask your assistant, from any chat app: "make a 20-second upbeat jingle for my bakery and send it here".

Sandboxed agents: if you run an agent in OpenClaw's sandbox, the skill runs inside the container, so python3 and L2C_API_KEY must exist there as well.

Models that suit OpenClaw

Model ID	Context	Takes images	$ / 1M in · out · cached	Now
`llama3.1` Llama 3.1 8B	128K	—	$0.016 · $0.032 · $0.008	resident
`ornith-1.5:9b` Ornith 1.5 9B (vision, 256K)	256K	yes	$0.050 · $0.080 · $0.020	resident
`ornith-1.5:35b` Ornith 1.5 35B (vision, 256K)	256K	yes	$0.18 · $0.30 · $0.050	on demand
`qwen3.8:27b` Qwen3.8 27B (vision + text)	128K	yes	$0.12 · $1.50 · $0.030	resident
`llama3.1:70b` Llama 3.1 70B	128K	—	$0.32 · $0.32 · $0.10	on demand
`deepseek-ocr:3b` DeepSeek-OCR 3B	128K	yes	$0.040 · $0.060 · $0.020	on demand
`granite4.1:3b` Granite 4.1 3B (text)	128K	—	$0.040 · $0.060 · $0.020	on demand

Tool calling and at least 64K context, from the live catalog. A model that is not resident answers 503 with "warming": true for a minute or two while it starts; the resident ones answer straight away.

Troubleshooting

The key shows up as ${L2C_API_KEY} in errors: the variable was not found. Check ~/.openclaw/.env and restart the gateway.
503 with "warming": true: the model is being loaded. Retry in a minute or use a resident model.
402: out of credit. Top up at /developers.
openclaw doctor --fix repairs a config that fails validation.