Claude visitors: new accounts receive $1.00 in bonus API credit within 30 days. One per person; usage credits only.

Agent Integration

OpenClaw + RunAPI

Connect OpenClaw to 250+ AI models through RunAPI. Generate video, images, music, and audio directly from your OpenClaw agent with one API key.

Get API Key View Documentation

Plug-and-play · No code changes · All modalities

Setup prompt

Configure this OpenClaw workspace to use RunAPI.

Requirements:
- Read the API key from RUNAPI_API_KEY. Do not hardcode the key.
- Add a model provider named runapi.
- Use baseUrl "https://runapi.ai/v1".
- Use api "openai-completions".
- Register provider-local model id "gpt-5.5" with display name "RunAPI GPT 5.5".
- Set the default model to "runapi/gpt-5.5" while keeping existing providers available.
- After editing config, run `openclaw models status` and confirm `runapi/gpt-5.5` appears.
- If RUNAPI_API_KEY is missing, ask me to set it before starting OpenClaw.

{
  agents: {
    defaults: {
      model: { primary: "runapi/gpt-5.5" },
      models: {
        "runapi/gpt-5.5": { alias: "RunAPI GPT 5.5" },
      },
    },
  },
  models: {
    mode: "merge",
    providers: {
      runapi: {
        baseUrl: "https://runapi.ai/v1",
        apiKey: "${RUNAPI_API_KEY}",
        api: "openai-completions",
        models: [
          {
            id: "gpt-5.5",
            name: "RunAPI GPT 5.5",
            reasoning: false,
            input: ["text"],
            contextWindow: 128000,
            maxTokens: 8192,
          },
        ],
      },
    },
  },
}

export RUNAPI_API_KEY="runapi_xxx"
openclaw models status
openclaw models set runapi/gpt-5.5
openclaw

Models Available via OpenClaw

HappyHorse/happyhorse-1.0-i2v

HappyHorse API access for text, image, and edit-video generation with 720p/1080p output and character-guided clips.

Qwen 2/qwen-2-edit-image

Qwen 2 API access for text-to-image and image editing from Alibaba's visual model family.

Producer/fuzz-0.8

Producer API access for FUZZ music generation from exact lyrics or instrumental production briefs.

ElevenLabs/text-to-dialogue-v3

ElevenLabs API access for text-to-speech and multi-speaker dialogue generation.

Qwen/qwen3.5-122b-a10b

Qwen API access via RunAPI: open-weight Qwen3.5, Qwen3.6, Qwen3-VL and Qwen3-Coder-Next language models on OpenAI-compatible Chat Completions.

Qwen/qwen3.5-27b

Qwen API access via RunAPI: open-weight Qwen3.5, Qwen3.6, Qwen3-VL and Qwen3-Coder-Next language models on OpenAI-compatible Chat Completions.

HOW IT WORKS

Connect in 3 Steps

Configure

Add your RunAPI API key to your OpenClaw configuration. One key accesses all models.
Select Models

Choose which models your agent can use. Filter by modality, quality, or cost.
Generate

Your OpenClaw agent generates media through RunAPI's unified API. Results are delivered via webhook or polling.

USE CASES

Automated Content

Let OpenClaw generate marketing videos, product images, and audio content on autopilot.

Multi-Modal Workflows

Chain video -> image -> music generation in a single agent conversation.

Cost-Optimized Routing

OpenClaw can compare model pricing and choose the most cost-effective option.

Integration questions

Can OpenClaw use RunAPI as a custom model provider?

Yes. Add RunAPI under models.providers with the /v1 base URL, then select its model.

Should the model ID include the provider prefix inside models.providers?

No. Inside models.providers.runapi.models, use the provider-local ID such as gpt-5.5. Use the provider-prefixed reference runapi/gpt-5.5 when selecting the model in OpenClaw defaults or commands.

Where should I store the RunAPI API key?

Store the key in RUNAPI_API_KEY and read it from the OpenClaw config. That keeps the credential out of version control and makes the same config usable across local shells, servers, and automation.

Can I keep my existing OpenClaw models as fallbacks?

Yes. Keep models.mode set to merge and add RunAPI as one provider. Your existing providers remain available, and you can add fallbacks under agents.defaults.model.fallbacks when you want recovery behavior.

Can OpenClaw generate images through RunAPI without installing extra skills?

Yes. The same RUNAPI_API_KEY and /v1 endpoint you configured for chat also accept image generation requests. Specify a model like flux-kontext-pro or imagen-4, and OpenClaw can call it through the existing RunAPI provider — no GitHub skill or separate API key needed. For a full guide, see the OpenClaw image generation page.

How does the OpenClaw agent know when an image or video task is ready?

Image, video, and music tasks are asynchronous. RunAPI returns a task ID on the first call. Your OpenClaw agent polls the status endpoint until the task completes, then receives the output URL. The RunAPI SDK handles polling automatically if you use it.

Other ways to run RunAPI from an agent

Workflow automation

Hermes Agent

Route generation steps through Hermes and let it pick the model per task. Same key, same balance, no extra setup.

Team assistant

WorkBuddy

Give a shared workspace assistant access to video, image, and LLM models with per-key spend limits.

SDK or MCP

Build your own

Wire any agent up with the RunAPI SDK or the MCP server — tools, polling, and webhooks included.

Building with a team?

We're here to help with enterprise setup, integrations, and technical questions.