Calls

Voice and video calls in which the agent and the customer each speak their own language and hear the other interpreted live.

A call has two sides, the agent and the customer. Each speaks their own language and hears the other interpreted into it, a couple of seconds behind, both ways at once. With video they also see each other.

Start a call#

Shell

curl https://api.spiiksi.fi/v1/calls \
  -H "Authorization: Bearer $SPIIKSI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent_language": "fi", "agent_name": "Acme Support",
       "customer_language": "es", "customer_name": "Lucía",
       "video": true, "external_id": "ticket-1234",
       "metadata": {"queue": "billing"}}'

JSON

{"id": "6f0c…", "status": "open", "livemode": true,
 "agent_name": "Acme Support", "agent_language": "fi",
 "customer_name": "Lucía", "customer_language": "es", "video": true,
 "external_id": "ticket-1234", "metadata": {"queue": "billing"},
 "agent_url": "https://spiiksi.fi/call/6f0c…#k=…",
 "customer_url": "https://spiiksi.fi/call/6f0c…#k=…",
 "created_at": "2026-10-08T09:00:00", "ended_at": null}

Field	Notes
`agent_language`	Required. The language the agent speaks and hears, such as `en`, `fi` or `pt-BR`.
`customer_language`	Leave it out and the customer chooses theirs when they open the link.
`video`	Whether the two may see each other. The voices are interpreted either way.
`external_id`, `metadata`	Your references, handed back with every webhook about the call.

The two links#

agent_url and customer_url open each side's call page. They are only in this answer: keep them, or make new ones later with POST /v1/calls/{id}/link.

Send customer_url to the customer: a text message, an email, your chat, your app.
Open agent_url for the agent, in a new tab or inside your own page with the JavaScript SDK.

Each link carries its side's key after #k=. The fragment isn't sent to servers or put in referrers. Anyone with a link speaks for that side, so treat links like passwords. A new link replaces the old one, and its microphone goes off the air.

The call page needs no account and no app: a browser with a microphone (and a camera, for video). The customer picks their language if it wasn't set, then starts a voice or video call. The call starts as soon as both sides are there.

During the call#

Interpretation runs both ways at once. Each side hears only the other's interpretation, never their original voice, with live captions.
Echo. On loudspeakers, the interpretation a person hears could reach their microphone. Unless they say they wear headphones, their microphone sends silence while the interpretation plays, so they speak in turns.
Video goes straight between the two browsers, without the voices. On strict networks it may not connect; the interpreted voice still does.
Quiet. A side that is silent for a minute stops being interpreted, and billed, until it speaks again.

End a call#

Shell

curl -X POST https://api.spiiksi.fi/v1/calls/$CALL_ID/end -H "Authorization: Bearer $SPIIKSI_KEY"

Both pages say the call has ended, the links stop working, and interpretation and billing stop. The agent can end it from their page too. A customer who leaves can come back with the same link until the call ends.

Find calls#

Shell

curl "https://api.spiiksi.fi/v1/calls?status=open&external_id=ticket-1234" \
  -H "Authorization: Bearer $SPIIKSI_KEY"

Calls are listed newest first, up to 100 at a time.

Other ways in#

A phone instead of a browser, for either side: Phone lines and SIP with AudioSocket.
Your own media server or contact-center platform playing a side: Joining a call from your system.
From Teams, Zendesk, Freshdesk, Salesforce or Genesys: Connectors.

Pricing#

A call is billed like live interpretation: per minute for each side while that side's interpretation runs, at your organization's prices. Languages served by the Default engine cost less than those that need the Plus engine. A quiet side pauses after a minute and isn't billed until it speaks again.