OpenClaw & MCP
Vydra is a native OpenClaw media provider for image, video, and speech generation. You can also connect Claude Desktop, Cursor, and other agents through the Model Context Protocol server.
MCP Setup
Add the following configuration to your MCP client settings:
{
"mcpServers": {
"vydra": {
"command": "npx",
"args": ["-y", "vydra-mcp"],
"env": {
"VYDRA_API_KEY": "your_key"
}
}
}
}API Key Required
OpenClaw Native Media Provider
OpenClaw includes the bundled vydra provider, enabled by default, so agents can call Vydra through image_generate, video_generate, and TTS without adding a separate MCP server. The provider reads VYDRA_API_KEY, supports image, video, and speech provider contracts, and uses the https://www.vydra.ai/api/v1 API base URL.
# Set the key for this shell/session
export VYDRA_API_KEY="vydra_live_..."
# Or use interactive OpenClaw onboarding
openclaw onboard --auth-choice vydra-api-key{
"agents": {
"defaults": {
"imageGenerationModel": {
"primary": "vydra/grok-imagine"
},
"videoGenerationModel": {
"primary": "vydra/veo3"
}
}
}
}image_generate(model="vydra/grok-imagine")
video_generate(model="vydra/veo3")
# tts uses the bundled speech provider: vydra/elevenlabs/tts| OpenClaw surface | Vydra models | Notes |
|---|---|---|
image_generate | vydra/grok-imagine | Text-to-image; one image per request. |
video_generate | vydra/veo3, vydra/kling | Veo3 supports text-to-video. Kling requires a remote reference image URL. |
tts | vydra/elevenlabs/tts | Returns MP3 audio with the default Rachel voice unless overridden. |
Which integration should I use?
vydra-mcp when you want portable Vydra tools in Claude Desktop, Cursor, or another MCP-compatible client.Available Tools
The Vydra MCP server exposes 13 tools that map to the core Vydra API surface.
| Tool | Description |
|---|---|
get_capabilities | List available workflows, models, and endpoints |
get_account | Get account info and credit balance |
generate_image | Generate an image from a text prompt |
generate_video | Generate a video from a text prompt or image |
lipsync_video | Sync lip movements in an image to an audio track |
swap_character | Swap a face from a source image onto a target image |
transfer_motion | Transfer motion from a video onto a static image |
generate_asmr | Generate ASMR audio content from a theme |
edit_image | Edit an existing image using a text prompt |
check_job | Check the status and progress of a job |
list_jobs | List recent jobs with optional status filter |
list_assets | List completed output assets |
get_asset | Get details and URL for a specific asset |
Usage Example
Once configured, you can use natural language in your MCP client:
"Generate a video of a cat playing piano"
"Check the status of my last job"
"Show me my recent assets"
"Generate an image of a sunset over mountains"
"Edit this image to add a rainbow in the sky"Troubleshooting
VYDRA_API_KEY not set / authentication errors
The MCP server reads your API key from the VYDRA_API_KEY environment variable. Make sure it is set in the env block of your MCP client config and that the key is valid (not revoked).
You can verify your key works by running: curl https://vydra.ai/api/v1/account -H "Authorization: Bearer YOUR_KEY"
npx cache issues / stale version
If the MCP server isn't picking up a new version, clear the npx cache: npx --yes clear-npx-cache, then restart your MCP client. The -y flag in the config args ensures the latest version is always fetched.
Connection refused / server not starting
Ensure Node.js v18+ is installed and available in your shell PATH. Try running npx -y vydra-mcp directly in a terminal to see startup errors. Check that no firewall or proxy is blocking outgoing HTTPS to vydra.ai.
Tool not found
If your MCP client doesn't list the Vydra tools, restart the client after saving the config. Some clients (e.g. Claude Desktop) require a full app restart to pick up new MCP server configurations.
402 / insufficient credits
The MCP server uses the same credit system as the REST API. Check your balance with the get_account tool or visit Billing to add credits.