callintel — Python SDK

Official Python SDK for the CallIntel voice agent platform. A thin, hand-written wrapper around httpx that covers the public org-side API.

Install

pip install callintel
# or
uv add callintel

Requires Python 3.10+.

Quickstart

from callintel import CallIntelClient

cli = CallIntelClient(
    api_key="ci_sk_...",           # or token="<jwt>"
    org_id="00000000-0000-...",
    # base_url="https://api.callintel.io",   # optional override
    # timeout=30.0,                          # optional
    # retries=3,                             # optional
)

agents = cli.agents.list()
print(agents)

call = cli.calls.place_phone_call(
    agent_id=agents[0]["id"],
    to_number="+15551234567",
)
print("Started call", call["call_id"])

Equivalent curl:

curl -X POST https://api.callintel.io/api/v1/orgs/$ORG_ID/calls/test-call \
  -H "X-API-Key: $CALLINTEL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"agent_id":"...","caller_phone":"+15551234567"}'

The client is a context manager:

with CallIntelClient(api_key="...", org_id="...") as cli:
    cli.agents.list()

Authentication

Pass exactly one of:

org_id is required and appears in every URL path.

Resources

Errors

from callintel import CallIntelAPIError, CallIntelClient

cli = CallIntelClient(api_key="...", org_id="...")
try:
    cli.agents.get("missing")
except CallIntelAPIError as exc:
    print(exc.status, exc.body, exc.request_id)

Retries + timeouts

Transient failures (408, 425, 429, 5xx) are retried with exponential backoff. Defaults: 3 retries, 30 s timeout. Override per-request via timeout= / retries= on client.request(...) if you need to.

Async client

import asyncio
from callintel import AsyncCallIntelClient

async def main() -> None:
    async with AsyncCallIntelClient(api_key="...", org_id="...") as c:
        agents = await c.agents.list()
        print(agents)

asyncio.run(main())

Mirrors the sync CallIntelClient 1:1 — same Agents, Calls, Campaigns, Webhooks, KnowledgeBases resources. Built on httpx.AsyncClient so it works inside any asyncio loop (FastAPI, Quart, Sanic, etc.).

Streaming live call events

from callintel import AsyncCallIntelClient, stream_call_events

async with AsyncCallIntelClient(api_key="...", org_id="...") as c:
    async for ev in stream_call_events(c, call_id="abc"):
        if ev["event"] == "transcript":
            print(ev["data"])
        if ev["event"] == "end":
            break

Wraps GET /api/v1/orgs/{org}/calls/{id}/live. Yields parsed {"event": str, "data": Any} dicts. Terminal events (end, timeout, not_found) close the iterator automatically.

Verifying webhooks

from fastapi import FastAPI, Request, HTTPException
from callintel import verify_webhook_signature

app = FastAPI()

@app.post("/callintel/webhook")
async def hook(request: Request) -> dict:
    body = await request.body()
    sig = request.headers.get("X-CallIntel-Signature")
    if not verify_webhook_signature(body, sig, "whsec_..."):
        raise HTTPException(status_code=401)
    # Trust the body now.
    return {"ok": True}

Constant-time (hmac.compare_digest) with 5-minute replay tolerance. Accepts both bytes and str payloads.

Not yet supported

• File-upload helpers for knowledge-base documents.

License

Apache-2.0