AI Integrations
Pluggable AI providers — including your own.
Thirdlane treats AI providers as configurable services, not embedded vendor relationships. Use the major commercial providers, run open-source models on-prem for compliance, or mix and match per service — transcription on one provider, sentiment on another, TTS on a third.
Commercial providers
Out-of-the-box integrations for the major commercial AI providers. Automatic model discovery means admins pick from a current model list per provider — no hand-maintained config.
OpenAI
GPT-class models for summarization, sentiment, action items, and entity extraction. Whisper for transcription. Automatic model discovery so admins pick from a current list.
Anthropic
Claude-class models for analysis tasks. Same configurable per-service options as other providers — pick which calls get summarization vs. QA vs. compliance.
Azure OpenAI
Azure-hosted GPT and Whisper deployments for tenants that need their AI traffic to stay within Microsoft’s compliance perimeter.
Google Cloud Speech
Cloud-hosted speech-to-text and text-to-speech for tenants standardizing on Google Cloud.
AWS
Amazon Transcribe for STT, Polly for TTS, and Bedrock for analysis — for tenants on AWS-aligned compliance.
Custom & on-prem providers
For deployments that can’t use third-party clouds, Custom Speech Services lets you run any TTS or transcription engine locally and route requests to it through the same configuration surface as commercial providers.
Custom Speech Services
Bring your own TTS or transcription engine. Drop in an executable at /usr/local/share/thirdlane/service/{tts,transcribe}; Thirdlane runs it as a child process per request.
Per-service environment
Each provider carries its own KEY=value environment file (one per line) so credentials and tuning knobs stay scoped to the service that uses them.
Air-gapped models
Run open-source STT/TTS models locally — Whisper.cpp, Coqui, Piper — and Thirdlane will route requests to them with the same configuration surface as the cloud providers.
Hybrid routing
Different services can route to different providers. Use one provider for transcription, another for sentiment, a third for TTS — without lock-in to a single vendor.
Why pluggable matters
Compliance & air-gap
Healthcare, finance, public sector — industries that can’t send call audio to a third-party cloud. Use Custom Speech Services to point at on-prem models.
Bring your own credit
Already paying Microsoft, Google, or AWS for AI? Plug those credentials in and use the credits you already have.
Service-provider economics
MSPs and UCaaS providers can choose a provider that fits their margin model — not be locked into the vendor’s preferred billing partner.
Best-of-breed analysis
Different models excel at different things. Use a fast/cheap model for summarization and a higher-tier model for QA scoring on a sampled subset.
What you get on top
Whichever provider(s) you wire in, the user-facing AI surface stays the same:
- Post-call summarization, sentiment, action items, QA scoring, categorization, entity extraction, and compliance checking — all in the consolidated recording detail window.
- Recording transcription with speaker turns, searchable from Reports.
- On-demand voice-message transcripts in Connect with copy-to-clipboard.
- Greeting transcription for non-TTS recordings.
- Visibility cues — call history avatar badges indicate which calls have been processed.
Configure AI for your customers.
We’ll walk through provider setup, per-service configuration, and the on-prem path for regulated workloads.