Connect Wavemaker to Cursor, Claude & Any AI Agent with MCP

If you live in an AI assistant — Cursor, Claude, Windsurf, VS Code Copilot — you can give it the ability to make video. Wavemaker exposes its full tool surface over the Model Context Protocol (MCP), so your agent can generate, review, refine, and render video as part of a conversation, with no API key to manage.

Why MCP (vs the REST API)

• MCP is for AI agents: one URL, OAuth sign-in, no key in your config. Best when a human or agent is driving interactively.
• REST is for code/CI/backends: an API key, deterministic calls. Best for automation (see Automate Video Creation with the API).

Both expose the same capabilities; pick the surface that matches how you’re working.

One-paste setup

The endpoint is https://api.wavemakr.com/mcp (streamable HTTP, OAuth 2.1 + PKCE). The first connection opens a browser sign-in.

Cursor

Create or merge into .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "wavemakr": { "url": "https://api.wavemakr.com/mcp" }
  }
}

Then approve it in Settings → Tools & MCP.

Claude Code

claude mcp add --transport http wavemakr https://api.wavemakr.com/mcp

Claude Desktop

Settings → Connectors → Add custom connector → paste https://api.wavemakr.com/mcp. On older versions, use the mcp-remote stdio shim in claude_desktop_config.json.

Windsurf

~/.codeium/windsurf/mcp_config.json:

{ "mcpServers": { "wavemakr": { "serverUrl": "https://api.wavemakr.com/mcp" } } }

VS Code (Copilot agent mode)

.vscode/mcp.json:

{ "servers": { "wavemakr": { "type": "http", "url": "https://api.wavemakr.com/mcp" } } }

Any other MCP client

Streamable-HTTP transport, URL https://api.wavemakr.com/mcp, OAuth 2.1 with PKCE (dynamic client registration supported). No pre-shared key.

What your agent can do once connected

• generate_video — one call runs the full pipeline: brief → script → storyboard → AI images/clips with native dialogue → voiceover → music → composed video.
• get_reviews — the per-scene quality verdicts the pipeline already computed (status, score, issues, recommended action + overall + consistency). Read-only, no credits.
• refine_video — natural-language edits to an existing video.
• render_video — returns an MP4 download link.
• Workspace tools — chain any of the 120+ internal tools (create_workspace → generate_image → review_image → generate_video_clip → compose_scenes → …) for custom pipelines. list_tools shows the catalog; describe_tool gives the schema.
• Libraries — list_assets, list_subjects + import_subject (reuse characters/products), ingest_asset (pull a public URL into your scope).
• predict_engagement — grade hook strength, retention risk, and per-platform fit before you post.
• repurpose_video — cut long-form video into captioned vertical clips.

Try it

Once connected, just ask:

“Generate a 30-second vertical ad for my product page at example.com, then score its engagement and tell me what to improve.”

Your agent will call generate_video, then predict_engagement, and report back — all in the conversation.

From agent to automation

MCP is perfect for interactive and agentic work. When you’re ready to run it on a schedule or wire it into your product, the same capabilities are available over REST — see Automate Video Creation with the API and the end-to-end generate→post pipeline.

See the developer docs →