Select an assistant to shape its voice and behavior, or create a new one to get started.
Assistant
Bound numbers
Calls to these phone numbers route to this assistant. Pick the telecom provider that owns the number.
Place an outbound call
Dial out from one of this assistant's bound numbers. Requires TELNYX_CONNECTION_ID set on the server.
Tools
The assistant can call these mid-conversation. Create tools in the Tools tab first, then check the ones to enable here.
Knowledge
Attached documents are injected into the system prompt every call. Create them in Knowledge.
Call logs
Call log records
Personas
Mock callers for testing your assistants. Each persona is a character prompt the caller model plays — an impatient regular, a confused first-timer — plus an optional voice for voice-mode runs. Voice fields left blank fall back to the server defaults.
Scenarios
What the caller tries to do, and the checks that decide pass/fail. Deterministic checks (tool_call, transcript, call) are compared in code; ai checks are extracted from the transcript by a judge model and compared with =, !=, >, <, >=, <=. See the API docs for the full shape.
Simulations
Pair an assistant with a persona and a scenario, then run it in chat mode (text through the real pipeline — fast and cheap) or voice mode (full audio through the browser-test path, with a recording).
Runs
Each run replays the scenario live against your assistant and grades the transcript. Expand a finished run for the per-check verdicts, the conversation, and (voice runs) the recording.
Add tool
Tools are HTTP endpoints the assistant can call mid-conversation (Vapi-style function calling). We POST signed JSON to your URL and speak back whatever you return. Verify X-APIFleet-Signature with the secret shown once on creation.
Add knowledge
Plain-text documents injected into the system prompt of any assistant they're attached to. Max 50 KB per document, 200 KB total per assistant. No embeddings — the whole content goes into the prompt.
Billing
Your balance, usage and payments. All amounts in USD.
Loading billing…
Prepaid balance
This internal workspace has no APIFleet charges and does not need a top-up. Your provider accounts continue to pay their own costs.
Available to spend
—
Held for calls
—
Total balance
—
Provider usage + $0.03 per connected AI minute. The platform fee is billed per second. No subscription or free credits.
Top up from $10. Calls using APIFleet provider keys need at least $1 available and start with a $1 hold. Holds extend as needed; unused funds return when the call finishes. Interrupted calls may keep a hold until their usage is confirmed.
Provider usage uses the recorded call usage and configured rates. When you bring an AI provider key, that provider bills you directly; APIFleet does not charge its usage again. Phone carriers, voice previews, file transcription and simulation caller/judge models currently use your own accounts. Manage credentials
Balance activity
Newest first
Date
Activity
Amount
Balance after
Loading activity…
Scroll the table to see amounts and balances.
Payments & receipts
Latest 100 payments
Date
Status
Wallet funds
Paid
Receipt
Loading payments…
Scroll the table to see payment amounts and receipts.
Provider credentials
Bring your own telecom and speech-provider accounts. Credentials are encrypted at rest (AES-256-GCM with the server's APIFLEET_ENCRYPTION_KEY) and never returned by the API once saved. Paid workspaces need their own provider credentials. Internal workspaces can use their configured server defaults.
Add webhook
We POST signed events to your URL. Verify X-APIFleet-Signature with the secret shown once on creation.
API keys
Use a key from your backend as Authorization: Bearer af_live_…. Shown once on creation.