Currai examples

Production-shaped examples for connecting AI applications to Currai with native HTTP capture.

Example	What it demonstrates
Chatbot	Streaming OpenAI chat with nested model and Exa MCP tool events
Vapi	Browser voice calls, finalized transcripts, and Vapi conversation traces
Retell AI	Retell web calls, rolling transcript reconciliation, and completed-call capture
OpenAI voice	Direct OpenAI Realtime WebRTC, live transcription, and nested voice-model traces

Each folder is an independent application. Enter the example you want to run, copy its environment template, install its dependencies, and start it:

cd openai
cp .env.example .env.local
pnpm install
pnpm dev

Read the example's own README for provider-specific credentials and verification steps.

Security

Real .env files are ignored. Only placeholder-only .env.example files are committed.
OpenAI, Retell AI, Vapi private, Exa, and Currai ingestion credentials stay in server code.
Browser-public or temporary provider credentials are identified explicitly in each example.
Currai capture is best-effort, redacts sensitive keys, uses short timeouts, and never changes the product request or voice-call result.
Generated dependencies, build output, coverage, local caches, and editor files are not committed.

Before deploying an example, apply your own authentication, authorization, rate limits, consent flow, data-retention policy, and abuse controls.

You can use the published Currai Skill to instrument another application with the same native HTTP capture pattern:

npx skills add https://github.com/curraiapp/skills --skill currai

See the Currai documentation for setup and provider-specific guides.