Deepgram MCP integration
Transcribes recorded audio into text, speaks answers back through text-to-speech, and reports project usage so spend stays visible.
8actions available
Three actions you can hand over today
Every action runs live through MCP. Nothing to build, nothing to maintain.
List deepgram projects
Lyro retrieves all your Deepgram projects, showing how voice AI workflows are organized for customer interactions. This provides visibility into your audio infrastructure.
Text-to-speech (rest)
Lyro converts written responses into natural-sounding speech via Deepgram, letting agents serve customers through voice channels with minimal effort. This powers voice-first support.
Transcribe pre-recorded audio
Lyro converts customer voice messages or call recordings into text that can be searched and responded to intelligently. This unlocks voice channel support instantly.
How businesses use Deepgram + Lyro
Each card is one request a support team gets, and the Deepgram actions Lyro runs to close it.
Check which Deepgram models fit the language you need
Lyro reads the public model catalogue and the metadata on a single model, so a question about supported languages, architecture, or model version is answered from Deepgram rather than a stale doc.
Get Public ModelsGet Model by IDList Think ModelsTurn a customer's audio file into readable text
When a customer sends a voicemail or a call recording, Lyro hands the file URL to Deepgram and returns the transcript, so the conversation continues in text without anyone replaying the audio.
Transcribe Pre-recorded AudioGet Public ModelsGet Model by IDReply out loud on a voice channel
Lyro sends the answer text to Deepgram's text-to-speech endpoint and gets natural speech back, so a customer on a voice channel hears the reply instead of reading it.
Text-to-Speech (REST)List Think ModelsGet Public ModelsWatch transcription usage before the bill arrives
Lyro pulls a project's audio duration, billable duration, and request counts and lists the projects and permission scopes tied to the key, so consumption is a question anyone can ask.
Get Project Usage SummaryList Deepgram ProjectsList Project Scopes
How it works
Get started in 3 steps
Connect once, then just ask. There is no workflow builder to learn and nothing to maintain — Lyro reads the Deepgram actions it has and picks the ones a request needs.
Connect Deepgram
Authorize the Deepgram account your team already uses — one consent screen, no API keys, no mapping tables. Lyro can only do what you granted that account, and you can disconnect it at any time.
Tell your agent what you need
Describe the job the way you would hand it to a teammate. Lyro maps it to the Deepgram actions that close it and chains as many as the request needs.
Watch it work
The agent runs the actions inside the conversation the customer is already in, so nobody copies data between tabs and your team can take over at any point.
Get started free
Deepgram · Lyro
What can you do in Deepgram without me setting anything up?
These, right away:
- Get model by ID
- Get public models
- List deepgram projects
Great, let's start with the first one.
FAQ
Everything else about Deepgram
Setup, permissions, and the limits of what Lyro can do inside Deepgram.
No. The connected toolkit exposes Transcribe Pre-recorded Audio, which takes a publicly accessible audio file URL, so transcription runs once a recording exists. Deepgram's streaming path is not part of the action set.
Every action available in Deepgram
All 8 actions your agent can call on Deepgram, straight from the live MCP connection.
Get model by ID
Retrieve metadata for a specific Deepgram model by its UUID.
Get public models
Retrieve metadata on all public Deepgram models (speech-to-text and text-to-speech).
List deepgram projects
List all Deepgram projects.
Get project usage summary
Retrieves aggregated usage statistics for a Deepgram project including total audio duration, billable duration, number of requests, channels processed, and confidence/relevance scores.
List project scopes
List all scopes for a specified Deepgram project.
List think models
List available think models for AI agent processing and voice agent configuration.
Transcribe pre-recorded audio
Transcribe pre-recorded audio files into text.
Text-to-speech (rest)
Convert text into natural-sounding speech.
The tools Deepgram sits next to
Same connection, same setup. Pick the next one your team already uses.
All images ai
Reports credit balance and key validity, lists automated and downloaded image batches, prunes old generations, and wires up print event webhooks.
Composio search
Answers off-catalog questions with live web, news, retail, travel, and maps results, then fetches the source page behind the answer.
Google cloud vision
Runs OCR and image annotation over pictures and PDFs, builds the Product Search catalogue, and polls long-running batch operations.
Metaphor
Fetches the parsed text and metadata for Metaphor documents by ID, several at a time, turning a set of result IDs into readable content.
Rev AI
Submits recordings for transcription, returns transcripts and SRT or VTT captions, and maintains the custom vocabulary behind accuracy.
Tisane
Detects the language of a message, flags abuse types and sentiment, scores text similarity, and paraphrases or translates a passage.
Ready to connect Deepgram?
Authorize the account and your agent has all 8 actions from the first conversation.