// MCP TOOL REFERENCE

FAL TOOLS

9 tools from the Fal MCP Server, categorised by risk level.

READ TOOLS

list-models List available fal.ai models from the built-in catalog. Filter by category: "image", "image_edit", "video", "audio", "3d", "utility". Empty for all.

WRITE TOOLS

edit-image Edit or transform an existing image using AI. Models: flux-kontext (default), nano-banana-2-edit (Gemini 3.1 Flash, fast), nano-banana-pro-edit (Ge... 2/5 generate-3d Generate a 3D model from an image. Models: meshy-v6, trellis-2 (fast) 2/5 generate-audio Generate audio: speech, music, or sound effects. Models: chatterbox-tts (TTS), minimax-speech (HD TTS), beatoven-music (music), beatoven-sfx (SFX) 2/5 generate-image Generate images from a text prompt. Models: flux-dev (balanced), flux-schnell (fast), flux-pro (best quality), recraft-v4 (design/typography), nano... 2/5 generate-video Generate a video from text and/or an image. Models: kling-v3-pro (best), sora-2-t2v, sora-2-i2v, ltx-2 (fast), wan-v2 2/5 upscale-image Upscale an image to higher resolution (4x by default) 2/5

DESTRUCTIVE TOOLS

remove-background Remove the background from an image 4/5

EXECUTE TOOLS

run-model Run any fal.ai model directly with custom arguments. Use for models not in the catalog or for advanced parameter control. 3/5

// FAQ

How many tools does the Fal MCP server have? +

The Fal MCP server exposes 9 tools across 4 categories: Read, Write, Destructive, Execute.

How do I enforce policies on Fal tools? +

Use Intercept, the open-source MCP proxy. Write YAML rules for each tool — rate limits, argument validation, or deny rules — then run Intercept in front of the Fal server.

What risk categories do Fal tools fall into? +

Fal tools are categorised as Read (1), Write (6), Destructive (1), Execute (1). Each category has a recommended default policy.

Enforce policies on Fal

Open source. One binary. Zero dependencies.

npx -y @policylayer/intercept

github.com/policylayer/intercept →