These are the tools the MOSS MCP server provides, grouped by capability. Naming one in a prompt helps when you want to steer a step, for example “Find three English female voices with list_voices.”
Your client fetches each tool's arguments from the server when it connects, and the agent fills them in from your request. Billing rules are in Sign-in and billing.
Image generation
| Tool | What it does | Billing |
|---|
generate_image | Generates one image from a text description and returns its link and actual width and height. You can pick the model (moss-image-base-1.0 or moss-image-sft-1.0) and the size (default 1024×1024; each side 256–2048 px and a multiple of 16, at most 2,097,152 pixels in total), add a negative prompt, turn on prompt enhancement, or customize or turn off the visible watermark. AIGC metadata is always kept. | Paid |
Speech generation
| Tool | What it does | Billing |
|---|
generate_speech | Reads text aloud in a given voice and returns a WAV audio link. Find voice IDs with list_voices. You can pick the model (moss-tts-1.5-flash or moss-tts-1.0-pro) and the speed (0.25–4, default 1); with moss-tts-1.5-flash you can also set the language. | Paid |
generate_speech_async | Same arguments as generate_speech, but returns a task ID right away; get the result with get_task. Use it for longer text. | Paid |
generate_dialogue | Turns a multi-speaker script into one audio file, each speaker in their own voice, and returns a WAV audio link. The model is always moss-ttsd-1.0. Good for podcasts and audio drama. | Paid |
generate_dialogue_async | Same arguments as generate_dialogue, but returns a task ID right away; get the result with get_task. Use it for longer scripts. | Paid |
Voices
| Tool | What it does | Billing |
|---|
list_voices | Finds voices: public voices, or your own (including ones you have made public). Search by name or tag. Returns each voice’s name, voice_id, language, gender, age range, and preview link, with paging when there are many results. | Free |
design_voice | Creates a new voice from a text description and returns a WAV audition. To keep the voice, ask to save it when generating; you then get a voice_id for speech and dialogue. Asking to save after the audition generates again, is charged again, and may sound different. | Paid |
design_voice_async | Returns a task ID right away; get the audition with get_task. It only makes an audition and cannot save the voice. | Paid |
Async tasks
_async tools return a task ID right away; check the result with get_task.
| Tool | What it does | Billing |
|---|
get_task | Checks an async task: processing, succeeded, failed, or canceled. On success it returns the result link; while running it may suggest when to check again. Failed or canceled tasks do not include a reason. | Free |
Account
Account tools are free.
| Tool | What it does | Billing |
|---|
get_usage | Shows available credits (net of frozen credits), frozen credits, credits used today and this month (Beijing time), the daily average over the last 7 days, and how many days your credits last at that rate. | Free |
create_mcp_key | Prepares the default API key that MCP uses for your account (named moss-mcp): reuses an existing one or creates it if missing. The key itself is never returned. | Free |