Fish Audio launches Fish Agent for no-code voice agents

Fish Audio launched Fish Agent, a no-code platform for building AI voice agents that make and take phone calls. Fish says the median time to deploy one is 32 seconds.

You start from a template or a short description, then set the agent's prompt and voice, attach a knowledge base of PDFs, Markdown files and URLs, and connect tools that call your own APIs during a conversation. Templates cover support, bookings, reservations, orders, research and reminders. Test the agent in the browser, then publish it to a website widget, your app, the API or a phone number. You can rent a number from Fish Audio or bring your own carrier, and an agent can transfer a call to a person with the context attached. Replies can come from a larger LLM billed at that model's token rate, or from your own OpenAI-compatible model endpoint.

Conversations can run on Live 3, the full-duplex voice model Fish Audio released two days earlier. It keeps listening while it speaks, so it can keep going through a quick mm-hmm, stop when you actually cut in, and wait while you finish a thought. Live 3 speaks English for now, and a cascaded pipeline is supported too.

Pricing is listed at $0.035 a minute through December 31, 2026, half the regular $0.07, and includes Live 3. Phone calls add $0.01 a minute and a dedicated number is $1.20 a month. You can start free with a $5 credit and no card, and enterprises can talk to sales. Fish Audio's docs describe Agents as a public beta, so features and APIs may still change.