Summary
Add pre-built tool/function definitions that allow AI agent frameworks (CrewAI, Anthropic tool use, OpenAI function calling, LlamaIndex) to call Deepgram's STT, TTS, and Audio Intelligence APIs as tools — without developers manually writing JSON schema definitions for each Deepgram capability.
Problem it solves
Developers building AI agents that need voice capabilities must manually write tool definitions (JSON schemas, function signatures, description strings) for every Deepgram API they want to expose. This is tedious, error-prone, and must be redone for each agent framework's tool format. Meanwhile, competitors ship pre-built "skills" packages (e.g., ElevenLabs' skills repo with 156 stars) that make their APIs instantly callable from any LLM agent. Pre-built tool definitions make Deepgram the default voice layer in the growing AI agent ecosystem.
Proposed API
from deepgram.tools import transcribe_audio, speak_text, analyze_audio
# Use with Anthropic tool use
tools = [transcribe_audio.to_anthropic(), speak_text.to_anthropic()]
# Use with OpenAI function calling
functions = [transcribe_audio.to_openai(), speak_text.to_openai()]
# Use with CrewAI
from crewai import Agent
agent = Agent(tools=[transcribe_audio.to_crewai(), analyze_audio.to_crewai()])
# Use with LangChain
from deepgram.tools import DeepgramTranscribeTool, DeepgramSpeakTool
tools = [DeepgramTranscribeTool(), DeepgramSpeakTool()]
# Each tool handles execution internally
result = await transcribe_audio.execute(url="https://example.com/audio.mp3")
Acceptance criteria
Raised by the DX intelligence system.
Summary
Add pre-built tool/function definitions that allow AI agent frameworks (CrewAI, Anthropic tool use, OpenAI function calling, LlamaIndex) to call Deepgram's STT, TTS, and Audio Intelligence APIs as tools — without developers manually writing JSON schema definitions for each Deepgram capability.
Problem it solves
Developers building AI agents that need voice capabilities must manually write tool definitions (JSON schemas, function signatures, description strings) for every Deepgram API they want to expose. This is tedious, error-prone, and must be redone for each agent framework's tool format. Meanwhile, competitors ship pre-built "skills" packages (e.g., ElevenLabs' skills repo with 156 stars) that make their APIs instantly callable from any LLM agent. Pre-built tool definitions make Deepgram the default voice layer in the growing AI agent ecosystem.
Proposed API
Acceptance criteria
Raised by the DX intelligence system.