Skip to navigation

OpenClaw

OpenClaw is an open-source agent platform that answers on Telegram, Discord, Slack, Signal, WhatsApp and more. Add HyperFlowSI as a custom OpenAI-compatible provider.

1

Set your key

export HYPERFLOW_API_KEY="sk-..."
2

Add the provider

Add HyperFlowSI to ~/.openclaw/openclaw.json:

openclaw.json
{
models: {
providers: {
hyperflow: {
baseUrl: "https://ai.hyperflow.fun/v1",
apiKey: "${HYPERFLOW_API_KEY}",
api: "openai-completions",
models: [
{
id: "openai/gpt-6-luna",
name: "GPT-6 Luna",
contextWindow: 1050000,
},
{
id: "deepseek/deepseek-flash",
name: "DeepSeek Flash",
},
],
},
},
},
agents: {
defaults: {
model: {
primary: "hyperflow/openai/gpt-6-luna",
fallbacks: ["hyperflow/deepseek/deepseek-flash"],
},
},
},
}

A model is referenced as hyperflow/ plus its HyperFlowSI id. fallbacks are tried in order if the primary model fails.

3

Start OpenClaw

openclaw gateway run

A model per agent

Agents can use different models, such as a cheaper one for the agent that answers a busy Telegram group. Set model under agents.entries.<agent id> to the same hyperflow/<model id> form, then route the channel to that agent with a bindings entry. See OpenClaw’s multi-agent routing.

Troubleshooting

  • No API key found: check echo $HYPERFLOW_API_KEY in the shell that starts OpenClaw.
  • 401: the key is wrong, deleted or disabled; check it on API keys.
  • 402: your balance is too low.
  • Model not found: the model must be listed under models.providers.hyperflow.models, and its id must match the Models page.