fal MCP 🔌 MCP server Freemium

What it is

The fal MCP server is fal.ai's official hosted endpoint at mcp.fal.ai/mcp. It gives an AI assistant access to fal's catalogue of more than 1,000 generative models (image, video, audio, 3D), plus fal's documentation, so the agent can find a model, read its input schema, check the price and run it.

Key features

• search_models and recommend_model to find a model for a task
• get_model_schema to read inputs and outputs before a call
• get_pricing to check cost before generating
• run_model for quick jobs and submit_job / check_job / get_job_result for long video renders
• upload_file to push reference images to fal's CDN
• search_docs for fal's own documentation

How to install

In Claude Code: claude mcp add --transport http fal-ai https://mcp.fal.ai/mcp --header "Authorization: Bearer YOUR_FAL_KEY". Cursor, Windsurf and ChatGPT/Codex can connect with OAuth instead. Any client that supports Streamable HTTP works. fal says your key is sent per request and not stored on the server.

Pricing and rights

Access to the server is free; you pay fal's standard per-output rates for each model run. There is no subscription. Example prices on fal's pricing page in September 2026: Seedream V4 at $0.03 per image, Kling 2.5 Turbo Pro at $0.07 per second of video, Veo 3 at $0.40 per second. Commercial rights follow each model's licence as listed on fal.

Who it is for

Creators and developers who want many models (FLUX, Kling, Veo, Seedream and more) behind one key, and agent builders who need to compare models by price.

Verdict

One of the broadest media servers available, and the pricing tool helps avoid surprise bills. You pay per output with no free tier stated, and model quality and licence vary, so you still need to pick models carefully.