MCP SERVER

Media generation for AI agents, behind a Model Context Protocol interface.

MediaEngine is an MCP server plus async REST API. It exposes intent-level tools (hero image, social teaser, email header, teaser video, article narration) and generic media tools so Claude, Cursor, Windsurf, Cline, and any MCP-compatible agent can queue image, video, and audio jobs and receive asset URLs back. Free tier: 25 credits, no credit card.

intent-level tools

One tool per editorial intent.

Each intent-specific tool picks the right aspect ratio, format, and provider for the surface, so the agent does not have to think about prompt engineering.

generate_article_hero_image

Hero banner for a blog post or article.

Takes a title and summary, picks a wide hero aspect ratio, and returns an asset URL when the job completes.

generate_social_teaser_image

Social media teaser (LinkedIn, Twitter/X, Instagram, Substack).

Auto-selects the aspect ratio for the platform and renders from the post content.

generate_email_header_image

Email header for newsletters and marketing emails.

Wide-format header generated from subject line and body text with tone and brand style hints.

generate_article_teaser_video

15–30 second teaser video from an article summary.

For YouTube Shorts, TikTok, and Instagram Reels. Async render, asset URL on completion.

generate_article_narration

Audio narration via text-to-speech.

Podcast-style narration, accessibility voiceovers, and audio newsletters. Multiple voices and playback speeds.

generic tools

Full provider surface, callable directly.

For agents that want explicit provider, model, and parameter control.

generate_image

Text-to-image with provider selection.

Inputs: prompt, size, model, style overrides. Providers: the MediaEngine generation pipeline, dalle3, stability-ai, chatgpt-image.

generate_video

Text-to-video with provider-specific parameters.

Inputs: inputParams (provider-specific), provider override. Providers: chatgpt-video (seedance, hailuo02, wan22, kling25, veo31, ovi).

generate_audio

Text-to-speech synthesis.

Inputs: text, voice, format (MP3, FLAC, AAC, Opus), playback speed. Provider: openai-tts.

wait_for_job

Block until a job completes; return the final asset URL.

Polling helper. Use when you want a single call instead of polling get_job_status yourself.

get_job_status

Check the current state of a job.

Returns queued, processing, done, or failed, plus the result URL when complete.

list_jobs

List recent media generation jobs.

Filter by status, type, and date. Sort and cursor-paginate.

cancel_job

Cancel a queued or in-flight job.

Marks the DB row cancelled. The provider may not honor cancel mid-render — the response indicates cancelled_running in that case.

install

Wire it into your agent in three steps.

Get an API key from the dashboard, set the environment variable, and point your MCP client at mcp-media-engine.

1. environment

Set the API key

MEDIAENGINE_API_KEY=me_live_...
# optional override
MEDIAENGINE_BASE_URL=https://api.mcpmediaengine.com
2. launch

Run via npx

npx -y mcp-media-engine
3. claude desktop

claude_desktop_config.json

Drop this into your Claude Desktop MCP config and restart.

{
  "mcpServers": {
    "mediaengine": {
      "command": "npx",
      "args": ["-y", "mcp-media-engine"],
      "env": {
        "MEDIAENGINE_API_KEY": "me_live_your_key_here"
      }
    }
  }
}
compatible agents

Works with the agents your team already runs.

The MCP server speaks stdio (npm) and streamable HTTP (hosted athttps://mcp.mcpmediaengine.com/mcp). The REST API is there for non-MCP integrations and direct HTTP calls.

  • Claude Desktop (native MCP support)
  • Claude Code (SDK support via .claude/claude.md)
  • Cursor (via MCP settings)
  • Windsurf (via AI settings -> MCP Servers)
  • Cline (via cline_config.json)
  • Codex CLI (OpenAI agent framework)
  • Gemini CLI (Google CLI agent)
  • Goose (agent framework)
  • Any MCP client over stdio
  • Direct REST API (HTTP, no MCP required)
delivery

Async by design

Submit a job, then either poll or receive a webhook. No long-lived blocking requests in the agent loop.

Webhook delivery

HMAC-signed completion callbacks

Batch mode

Up to 500 jobs / request

Agent-native

Designed around MCP semantics

Free tier

25 credits, no card

Plug MediaEngine into your agent today.

25 free credits. No credit card. Cancel any time.