Skip to main content
Unifically LogoUnificAlly

Connect any agent to every generative model

Add one endpoint and Claude, Cursor, or any MCP client can make video, images, music, and speech in chat. 90+ models, one API key.

Browse the tools
mcp.unifically.com/mcp

Works with claude.ai, Claude Code, Claude Desktop, Cursor, VS Code, OpenClaw, and any other MCP client.

mcp.json
{
  "mcpServers": {
    "unifically": {
      "url": "https://mcp.unifically.com/mcp",
      "headers": { "Authorization": "Bearer sk-..." }
    }
  }
}

https://mcp.unifically.com/mcp

What is MCP

The open standard for assistant tools

MCP (Model Context Protocol) is the open standard that lets AI assistants like Claude, Cursor, and ChatGPT call outside tools. Add the Unifically server once and your assistant can make media in chat. Ask for a 9:16 video of a sunset over Tokyo and the finished file comes back in the conversation.

Every request carries your own API key, so usage is billed to your account only. No account linking, no OAuth. The key is the whole setup.

Transport

Streamable HTTP

Models

90+

MCP access

Free

Auth

Your API key

Setup

Connect your client

Pick your client and paste the snippet. The only credential you need is an API key from your dashboard.

Settings → Connectors → Add custom connector, then paste this URL:

connector url
https://mcp.unifically.com/mcp?api_key=sk-...

Legacy SSE clients can connect via https://mcp.unifically.com/sse instead.

Authentication

Three ways to send your key, in order of preference:

Authorization: Bearer sk-...

Recommended. Works in any client that can set headers.

X-API-Key: sk-...

Alternative header, same effect.

?api_key=sk-...

For clients that only take a URL, like claude.ai connectors. It works, but URLs can end up in logs, so use a header when you can.

Tools

What your assistant can do

Fourteen tools: generation, editing, upscaling, file uploads, the model catalog, and cost control.

generate_video

Text and image to video with Veo (default), Kling, Hailuo, Seedance, and Grok. Takes start/end frames and reference images.

prompt · image_url · duration · aspect_ratio

generate_image

Text-to-image and editing with Nano Banana (default), Flux, GPT Image, Qwen, SeeDream, and Grok.

prompt · image_urls · resolution · aspect_ratio

generate_music

Full songs with Suno. Describe the track or bring your own lyrics.

prompt · mode · instrumental · audio_url

generate_audio

Text-to-speech, sound effects, multi-voice dialogue, and speech-to-text transcription.

model · prompt · voice_id · audio_url

edit_video

Edit an existing clip from a prompt — restyle, add or remove elements — or transfer motion from a video onto a subject image.

video_url · prompt · image_urls

upscale_media

Upscale any image (2x/4x) or video (up to 8K) with Topaz. Veo and Grok clips can also use their native upscalers.

image_url · video_url · upscale_factor

extend_video

Continue a finished video. Veo adds 8 seconds; Grok takes preset or custom lengths.

task_id · model · prompt

upload_file

Turns a local file into a CDN URL the models accept as input.

file_path · url · base64

list_models / list_resources

Browse the live model catalog and a provider's voices — ElevenLabs, Kling, and Veo.

category · provider · resource

list_tasks / get_task

Find past generations, resume them, or recover a result after a timeout.

task_id · status · page

dry_run_cost / check_balance

Price a job before running it and check your remaining credit.

model · input

In practice

One request, one finished file

Ask in plain language. The assistant picks the tool, quotes the price, and returns the result URL in the same conversation.

  1. 1

    The assistant can price the job first with dry_run_cost and ask before spending.

  2. 2

    Quick jobs return the finished URL in one call. Long jobs return a task_id right away and the assistant polls get_task until done.

  3. 3

    Every result reports what it cost, billed to your key at API prices.

claude · unifically
› "make a 9:16 video of a sunset over Tokyo"⚙ dry_run_cost · google/veo-3.1-fast · duration=8  est $0.40⚙ generate_video · prompt="sunset over Tokyo" · aspect_ratio=9:16  → task_9f27c1 · processing⚙ get_task · task_id=task_9f27c1✓ https://files.unifically.com/v/sunset-tokyo-9f27.mp4 · $0.40

Design

How it works

One call, one result

Ask for an image and the assistant gets the finished URL back in the same tool call. Long jobs like high-quality video and full songs return a task_id right away, and the assistant polls get_task until they finish. Nothing gets lost, even across proxy timeouts.

Your files

Media inputs are public URLs. The assistant uploads local files through upload_file, or straight to files.unifically.com from a shell, and gets a CDN URL back.

Cost transparency

Every result reports what it cost. The assistant can quote a price with dry_run_cost before spending, and the server's built-in guidance tells it to ask before pricey jobs. It never chains paid steps you didn't request.

Errors that explain themselves

Every failure tells the assistant what to fix: a wrong input, provider capacity, or (only on a real 402) an actual balance problem.

Works with OpenClaw

Running OpenClaw? Point it here.

OpenClaw is a free, open-source personal AI agent that runs on your own computer and answers from WhatsApp, Telegram, Discord, Slack, and other chat apps. Connect it to the Unifically MCP server and it can reply with finished videos, images, and songs, not just text. Our OpenClaw guide covers install and setup end to end.

MCP access is free

You pay only for generations, at the same per-task prices as the API, billed to your key. Check any model's price on the pricing page, or ask your assistant to dry-run the job first.

See model pricing

FAQ

Frequently asked questions

A hosted Model Context Protocol server at mcp.unifically.com. It gives AI assistants tools for video, images, music, and speech across 90+ models, including Veo, Kling, Nano Banana, Flux, Suno, ElevenLabs, and Topaz, with one API key.

Any MCP client. claude.ai, Claude Code, Claude Desktop, Cursor, and VS Code get copy-paste snippets. The server speaks Streamable HTTP and keeps a legacy SSE endpoint for older clients.

MCP access itself is free. You pay only for generations, at the same per-task prices as the Unifically API, billed to the API key each request carries.

No. Every request carries your API key, in an Authorization header or, for URL-only clients, as an api_key query parameter. The key is the entire setup.

Yes. The upload_file tool turns a local file into a CDN URL that models accept as input, and you can pass any public image, video, or audio URL directly.

On claude.ai, go to Settings, then Connectors, and add https://mcp.unifically.com/mcp?api_key=sk-... as a custom connector. In Claude Code, run claude mcp add --transport http unifically https://mcp.unifically.com/mcp with an Authorization header. Get an API key first; the exact snippets are in the setup section above.