Anthropic Claude — available as an alternative LLM powering your Fluents voice agents.
Fluents' voice AI stack runs on Google Gemini as its default conversation engine. For enterprise customers who have specific model requirements — existing Anthropic agreements, compliance mandates around model provider, or a preference for Claude's specific reasoning style — Claude can be configured as the LLM that powers your agents instead.
The rest of the stack stays the same: ElevenLabs for voice synthesis, Deepgram for transcription, and Fluents Insights for structured output. Only the conversation engine changes.
Claude powers the conversation layer of your Fluents agents — handling intent recognition, context tracking, and response generation on every call
Ideal for organizations with existing Anthropic enterprise agreements or compliance requirements tied to specific model providers
Claude's careful instruction-following makes it a strong fit for sensitive domains like legal intake and healthcare where agents must stay precisely within defined boundaries
How Fluents' AI Stack Works
Every call Fluents handles passes through three layers: Deepgram converts speech to text, the conversation engine (Gemini by default) processes the transcript and generates a response, and ElevenLabs synthesizes that response into natural speech. Model selection means choosing which LLM sits in that middle layer — everything else stays identical.
Legal Intake: Agents That Stay Within Strict Boundaries
Legal intake agents must never offer legal advice, must stay on their qualification script, and must escalate to a human attorney at the right moment. Claude's instruction-following and constitutional AI training make it exceptionally good at holding to defined boundaries even when callers push against them. A personal injury firm running intake calls can trust that the agent won't improvise outside its scope.
Healthcare: Sensitive Conversations, Careful Responses
Patient-facing agents that handle appointment scheduling, post-discharge check-ins, or billing inquiries need to be helpful without crossing into medical advice. Claude's training aligns well with the careful, bounded helpfulness these conversations require — acknowledging patient concerns without overstepping into clinical territory.
Enterprise: Existing Anthropic Agreements
Organizations that have negotiated enterprise data processing agreements with Anthropic can configure Fluents to route conversation engine calls through those agreements — keeping all AI processing under a single vendor relationship with known terms.