API access

What you can build with it

This is a voice agent API, not a raw model API. You build agents (prompt, voices, tier, tools) and place calls; the engine runs the speech, model and voice pipeline.

Browser calls send 16 kHz PCM16 audio up and receive 24 kHz down, with live transcripts. Speech-to-text takes PCM16, mono, 16 kHz only.

One key, kept on your server

One Voice API key works on both voice hosts: as a bearer token on REST, and on the call socket. Speech-to-text uses its own token from the Speech to Text page.

export ALEBEX_API_KEY="wt_..."   # server side only curl https://api.alebex.ai/api/v1/public/agents \  --header "Authorization: Bearer $ALEBEX_API_KEY"

• One key per account. It doesn’t expire and has no scopes.
• Rotating it deletes the old key at once. You can revoke it anytime.
• Each key is limited to 120 reads and 60 writes a minute; past that, requests return 429 RATE_LIMITED.
• Browser calls open the socket with your key, so proxy the socket through your backend rather than putting the key in a page.

curl https://api.voice.alebex.ai/public/call/phone \  --header "Authorization: Bearer $ALEBEX_API_KEY" \  --header "Content-Type: application/json" \  --data '{"agentId": "agt_...", "to": "+16045550142",           "twilio": {"accountSid": "AC...", "authToken": "...",                      "phoneNumber": "+16045550100"}}' # 201 {"id": "...", "providerCallId": "...", "status": "queued"}

A signed report when every call ends

Each call’s report goes to your endpoint when it ends: transcript, recording, outcome, tool calls and cost.

• Up to 10 endpoints, each with the report fields you choose (25 available).
• Optional HMAC-SHA256 signing, sent in the X-Alebex-Signature header.
• Three delivery attempts with backoff, a delivery history, and one-click resend.