Skip to main content

FlyMy MCP

Gives an agent access to the full FlyMyAI model catalog through the local FlyMy MCP gateway (Streamable HTTP, JSON-RPC). The tool list is dynamic - it is discovered at runtime via tools/list, and calls are dispatched via tools/call. Generation runs go to api.flymy.ai under your own account key, so the user who triggers the run is the one billed.

The capabilities are model-driven: instead of a fixed set of generators, the agent first picks a model with recommend_model, reads its parameter schema with get_model_schema, then launches it with run_model.

What it can do​

MethodWhat it does
recommend_modelPick the best FlyMyAI model for a task (text-to-image, image-to-image, text-to-video, image-to-video, and more). Returns ranked model_id candidates - the source of truth for what models exist. Discovery only, no charge.
get_model_schemaReturn the input parameter schema for a given model_id so the agent knows exactly which arguments to pass. Discovery only, no charge.
run_modelRun a chosen model_id with parameters from its schema (e.g. nano-banana, seedream, seedance, veo, flux, imagen). Returns an agent_file with id and public_url for the generated image or video.
analyze_imageDescribe or analyze an image with a VLM before generating - useful to ground an edit or image-to-video on the actual source.

Discovery tools (list_* / get_*) are listed dynamically by the gateway and may expand over time; treat recommend_model as the authoritative list of available models rather than relying on fixed names.

Looking for the Models API?

The methods above are agent-facing MCP actions. For direct programmatic access to all FlyMyAI models - including schema lookup (client.schema()), async runs (client.async_run()), and the full catalog - see the Models API and the Python / JavaScript SDKs.

How to get credentials​

None - FlyMy MCP is a built-in tool. It runs against the local FlyMy MCP gateway and bills generation runs to your existing FlyMyAI account, so there is no extra key to paste. Just enable it.

Fields to fill in FlyMyAI​

None - the MCP uses your FlyMyAI account credentials.

Troubleshooting​

  • No model fits the task - always call recommend_model first; never refuse a generation citing "no capability" before checking, since the catalog is the only source of truth and changes over time.
  • Parameter error on run - re-read the schema with get_model_schema for that model_id and pass arguments exactly as specified.
  • Run needs your account key - generation is billed to your FlyMyAI account; if no account API key is set, a run fails loudly rather than charging a shared key.