Fish Audio

@fish-audio list all my agents and show me which ones are still drafts versus published

Found 4 agents in your workspace. Two are published: 'Customer Support Bot' (ID: agt_abc123, voice: Emily-Professional) and 'Sales Demo Assistant' (ID: agt_def456, voice: Marcus-Friendly). Two remain drafts: 'Internal FAQ Agent' (ID: agt_ghi789, no voice assigned yet) and 'Product Onboarding Guide' (ID: agt_jkl012, voice: Sarah-Warm, prompt configured but not published). Draft agents won't appear in your public API until you explicitly publish them.

This example demonstrates the List Agents tool with workspace-wide visibility. Useful for auditing which conversational agents exist and their publication state. Requires API_KEY auth with read access to the workspace; the tool returns metadata only—you'll need Get Agent to inspect full conversation configuration or prompts.

@fish-audio update my 'Customer Support Bot' agent to use a more empathetic opening line in its prompt

Updated agent agt_abc123 ('Customer Support Bot'). The draft prompt now begins: 'I'm here to help you resolve any issue—let's work through this together.' The agent's voice (Emily-Professional) and core conversation settings remain unchanged. This update is saved as a draft; the live published version still uses the original prompt until you explicitly publish the new draft.

Showcases the Configure Agent tool's ability to iteratively refine conversation behavior without disrupting the live agent. Important side-effect: changes are draft-only until published, so the public API continues serving the old version. If you need the update live immediately, you must follow up with a separate publish action (not shown here).

@fish-audio check my account balance and tell me if I have enough quota to run 500 synthesis requests this month

Your Fish Audio account shows 12,000 characters remaining in your free package allowance and $8.47 in API credits. A typical synthesis request consumes roughly 150–300 characters depending on response length. At 200 characters per request, 500 requests would require ~100,000 characters. Your current free quota covers about 60 requests; the remaining 440 would draw from your $8.47 credit, which may not be sufficient depending on your plan's per-character rate. Consider topping up credits or upgrading your package before launching a high-volume campaign.

Illustrates the Get Account Balance tool paired with the AI's reasoning to translate raw quota into actionable capacity planning. Helpful for avoiding mid-campaign quota exhaustion. The example assumes typical character-per-request averages; actual consumption varies by agent verbosity and conversation length. Rate limits and billing details live in Fish Audio's dashboard, not the MCP response.

When Fish Audio fits a small support team's IVR needs

A 6-person support team wants a voice agent that handles tier-1 questions before escalating to Slack. Fish Audio's agent-creation and voice-model tools let you draft, test, and publish conversational flows without leaving Switchy. The Configure Agent tool means your team can iterate on prompt tone and voice selection in real time during a standup, then push live once the wording feels right. The catch: if your support queue already lives in Zendesk or Intercom, you'll need custom webhooks to bridge session data back into your ticketing system—Fish Audio doesn't ship those connectors. Best fit if you're building a net-new voice layer and your team is comfortable wiring the last mile. If you need plug-and-play CRM integration, look at a platform MCP instead.

Why a two-person content team should skip this MCP

A podcaster and editor want to generate custom intro voiceovers each week without hiring talent. Fish Audio's Get Voice Model and synthesis capabilities sound perfect, but the MCP exposes agent-management tools—not direct text-to-speech endpoints. You'd create an agent, configure its voice, then trigger a session just to render 15 seconds of audio. That's three API calls and session cleanup when a dedicated TTS MCP (ElevenLabs, Play.ht) does it in one. The List Agent Sessions tool is overkill for a workflow that never needs conversation history. Use Fish Audio if you're building a conversational experience that happens to include scripted segments. For pure asset generation at small scale, a simpler TTS integration saves you the agent-lifecycle overhead.

When Fish Audio scales a 10-person sales team's demo library

A sales team runs discovery calls in English, Spanish, and Mandarin, and wants on-demand voice demos that adapt to each prospect's vertical. Fish Audio's agent-draft workflow and voice-model library let you maintain three language-specific agents in Switchy, each with a tailored prompt and regional voice. The Get Account Balance tool surfaces quota burn before a big campaign, so your ops lead knows when to top up API credits. The List Agents filter helps your team archive old verticals (fintech, healthcare) without cluttering the active roster. Trade-off: if your demo scripts change weekly, the draft-publish cycle adds friction—you're editing agent config, not just swapping a text file. Best when your messaging is stable enough to version as agents, and you need multi-language coverage a simple TTS tool can't match.