twilio-call-recordings
$
npx mdskill add openai/plugins/twilio-call-recordingsRecord Twilio voice calls correctly using the right methods for compliance and quality assurance
- Solves incorrect use of Twilio's Record verb versus Dial record for call recording
- Leverages Twilio Voice API, REST API, and Conference recording features
- Analyzes call context to recommend dual-channel, pause, or conference recording
- Delivers recordings via status callbacks and provides tools for compliance and QA
SKILL.md
.github/skills/twilio-call-recordingsView on GitHub ↗
---
name: twilio-call-recordings
description: >
Record Twilio voice calls correctly. Covers the critical distinction
between Record verb (voicemail) and Dial record (call recording),
dual-channel for QA, mid-call pause for PCI, Conference recording, and
the ConversationRelay workaround. Use this skill whenever you need to
capture call audio for compliance, QA, or analytics.
---
## Overview
Twilio offers multiple recording methods. Choosing the wrong one is the **#1 developer mistake** in voice — using `<Record>` when you mean `<Dial record>` produces voicemail behavior instead of call recording.
| Method | What it does | Use when |
|--------|-------------|----------|
| `<Record>` verb | Records the CALLER only (voicemail-style) | Leaving a message, capturing input |
| `<Dial record>` | Records BOTH parties on a call | Call recording for two-party calls |
| `<Start><Recording>` | Starts a recording alongside other verbs | ConversationRelay, multi-verb flows |
| Conference `record` | Records the conference mix | Multi-party calls |
| Recordings REST API | Programmatic control mid-call | Pause during payment (PCI) |
---
## Prerequisites
- Twilio account with a voice-capable phone number — see `twilio-account-setup`
- `TWILIO_ACCOUNT_SID` and `TWILIO_AUTH_TOKEN` — see `twilio-iam-auth-setup`
- SDK: `pip install twilio` / `npm install twilio`
- A webhook endpoint for recording status callbacks
- **Compliance check:** Recording consent requirements vary by jurisdiction — see `twilio-compliance-traffic`
---
## Quickstart
### Record a Two-Party Call (Most Common)
Use `<Dial record>` — NOT `<Record>`.
**Python (Flask)**
```python
from flask import Flask, request
from twilio.twiml.voice_response import VoiceResponse
app = Flask(__name__)
@app.route("/voice", methods=["POST"])
def incoming_call():
response = VoiceResponse()
response.say("This call may be recorded for quality assurance.")
dial = response.dial(
record="record-from-answer-dual", # dual-channel: agent on one, caller on other
recording_status_callback="https://yourapp.com/recording-status"
)
dial.number("+15558675310") # agent's phone
return str(response)
```
**Node.js (Express)**
```node
app.post("/voice", (req, res) => {
const response = new VoiceResponse();
response.say("This call may be recorded for quality assurance.");
const dial = response.dial({
record: "record-from-answer-dual",
recordingStatusCallback: "https://yourapp.com/recording-status",
});
dial.number("+15558675310");
res.type("text/xml").send(response.toString());
});
```
### Handle the Recording Status Callback
> **Security:** Validate `X-Twilio-Signature` on recording callbacks in production. Without validation, attackers could POST fake recording URLs to your endpoint.
**Python (Flask)**
```python
@app.route("/recording-status", methods=["POST"])
def recording_status():
recording_sid = request.form["RecordingSid"]
recording_url = request.form["RecordingUrl"]
call_sid = request.form["CallSid"]
status = request.form["RecordingStatus"] # "completed", "failed"
duration = request.form.get("RecordingDuration", 0)
if status == "completed":
# Store recording reference
save_recording(call_sid, recording_sid, recording_url, duration)
return "", 200
```
---
## Key Patterns
### Recording Modes for `<Dial record>`
| Mode | What's recorded | Use case |
|------|----------------|----------|
| `record-from-answer` | Single channel, both parties mixed | Simple recording |
| `record-from-answer-dual` | Dual channel — caller on left, agent on right | QA (separate agent/caller audio) |
| `record-from-ringing` | Records from ring, not answer | Capture ring time + full call |
| `record-from-ringing-dual` | Dual channel from ring | QA with ring time |
**Always use `dual` for QA and analytics.** Dual-channel lets speech analytics tools (like Conversation Intelligence) distinguish agent from caller.
### Conference Recording
Record multi-party calls via the Conference:
**Python**
```python
response = VoiceResponse()
dial = response.dial()
dial.conference(
"support-room-123",
record="record-from-start", # Records from when conference starts
recording_status_callback="https://yourapp.com/conf-recording-status"
)
```
**Note:** Conference recording captures the main audio mix. Coach/whisper audio is NOT included. See `twilio-conference-calls`.
### ConversationRelay Recording
**Critical:** `record:true` on the REST API call is **silently ignored** with ConversationRelay. No error. No recording.
**Correct approach:** Use `<Start><Recording>` in TwiML before `<Connect>`:
**Python**
```python
@app.route("/voice", methods=["POST"])
def voice():
response = VoiceResponse()
response.say("This call may be recorded.")
# Start recording BEFORE connecting ConversationRelay
start = Start()
start.recording(
recording_status_callback="https://yourapp.com/recording-status",
recording_status_callback_event="completed"
)
response.append(start)
# Now connect ConversationRelay
connect = Connect()
connect.conversation_relay(url="wss://yourapp.com/ws/voice")
response.append(connect)
return str(response)
```
**Node.js**
```node
app.post("/voice", (req, res) => {
const response = new VoiceResponse();
response.say("This call may be recorded.");
const start = response.start();
start.recording({
recordingStatusCallback: "https://yourapp.com/recording-status",
recordingStatusCallbackEvent: "completed",
});
const connect = response.connect();
connect.conversationRelay({ url: "wss://yourapp.com/ws/voice" });
res.type("text/xml").send(response.toString());
});
```
### Mid-Call Pause for PCI Compliance
Pause recording when a customer provides payment information:
**Python**
```python
def pause_recording_for_payment(call_sid, recording_sid):
"""Pause recording during credit card capture."""
client.calls(call_sid).recordings(recording_sid).update(
status="paused"
)
def resume_recording(call_sid, recording_sid):
"""Resume recording after payment processed."""
client.calls(call_sid).recordings(recording_sid).update(
status="in-progress"
)
```
**Node.js**
```node
async function pauseForPayment(callSid, recordingSid) {
await client.calls(callSid).recordings(recordingSid).update({
status: "paused",
});
}
async function resumeRecording(callSid, recordingSid) {
await client.calls(callSid).recordings(recordingSid).update({
status: "in-progress",
});
}
```
**PCI DSS:** Never record card numbers. Use Twilio's `<Pay>` verb when possible. If collecting verbally, pause recording for the duration. PCI Mode is IRREVERSIBLE and account-wide — use a sub-account if only some calls need PCI.
### Accessing Recordings
**Python**
```python
# List recordings for a specific call
recordings = client.recordings.list(call_sid=call_sid)
for recording in recordings:
print(f"SID: {recording.sid}")
print(f"Duration: {recording.duration}s")
print(f"URL: https://api.twilio.com{recording.uri.replace('.json', '.mp3')}")
# Download a recording
import requests as req
audio = req.get(
f"https://api.twilio.com/2010-04-01/Accounts/{account_sid}/Recordings/{recording_sid}.mp3",
auth=(account_sid, auth_token)
)
with open("recording.mp3", "wb") as f:
f.write(audio.content)
# Delete a recording (GDPR right to deletion)
client.recordings(recording_sid).delete()
```
### Recording Storage & Retention
| Feature | Default | Notes |
|---------|---------|-------|
| Storage location | Twilio cloud | Can configure external storage (S3, GCS) |
| Retention | Indefinite | Delete manually via API or set auto-delete policy |
| Formats | WAV (default), MP3 | Request MP3 by appending `.mp3` to URL |
| Encryption | At rest | Additional encryption with PCI Mode |
---
## Common Errors
| Symptom | Cause | Fix |
|---------|-------|-----|
| Recording captures only caller (no agent) | Used `<Record>` verb instead of `<Dial record>` | Switch to `<Dial record="record-from-answer">` |
| No recording at all | Used REST API `record:true` with ConversationRelay | Use `<Start><Recording>` in TwiML |
| Recording is empty / silent | Webhook endpoint unreachable, recording never started | Check StatusCallback URL reachability |
| Recording has both parties on same channel | Used `record-from-answer` (mono) | Use `record-from-answer-dual` for separate channels |
| Coach audio missing from conference recording | Expected behavior — coach audio isn't in the mix | Record coach's call leg separately |
---
## CANNOT
- **`recordingTrack` has no observable effect via TwiML** — The `<Start><Recording>` TwiML parameter `recordingTrack` does not isolate tracks. Use the Recordings REST API with `recordingTrack` for actual track isolation.
- **Cannot start API recordings on ConversationRelay calls** — REST API `record:true` is silently ignored ("not eligible for recording"). Must use `<Start><Recording>` before `<Connect>` in TwiML.
- **Cannot pause/resume recordings via TwiML** — Only available via the REST API (`update` with `status="paused"` or `status="in-progress"`).
- **Cannot get dual-channel conference recordings** — Conference recording is always mono (mixed).
- **Cannot get dual-channel from Calls API without explicit param** — `Record=true` defaults to mono. Must specify `recordingChannels: 'dual'`.
- **Cannot transcribe PCI-mode recordings** — Recordings created while PCI mode was enabled cannot be transcribed, even after PCI is disabled.
- **Cannot use `<Record>` verb for call recording** — `<Record>` captures the caller only (voicemail-style). Use `<Dial record>` or `<Start><Recording>` for call recording.
- **Cannot capture coach/whisper audio in conference recordings** — Supervisor whisper is excluded from the mix
- **Cannot reverse PCI Mode** — PCI Mode is irreversible and account-wide. Once enabled, all recordings are encrypted.
- **Cannot auto-delete recordings without configuration** — Recordings are retained indefinitely unless you configure auto-deletion
- **Cannot avoid larger file sizes with dual-channel** — Dual-channel recordings are ~2x the size of mono. Factor into storage costs.
---
## Next Steps
- **Conference calls:** `twilio-conference-calls`
- **Agent routing:** `twilio-taskrouter-routing`
- **Compliance:** `twilio-compliance-traffic`
- **Debug recording issues:** `twilio-debugging-observability`
More from openai/plugins
- accessibility-and-inclusive-visualizationMake data visualizations accessible and inclusive. Use when the user needs chart or diagram accessibility guidance, text alternatives for complex visuals, color and contrast review, keyboard support, reduced-motion behavior for animation or parallax, or an accessibility QA workflow for exported figures, UML-like diagrams, and dashboards.
- agent-browserBrowser automation CLI for AI agents. Use when the user needs to interact with websites, verify dev server output, test web apps, navigate pages, fill forms, click buttons, take screenshots, extract data, or automate any browser task. Also triggers when a dev server starts so you can verify it visually.
- agent-browser-verifyAutomated browser verification for dev servers. Triggers when a dev server starts to run a visual gut-check with agent-browser — verifies the page loads, checks for console errors, validates key UI elements, and reports pass/fail before continuing.
- agents-sdkBuild AI agents on Cloudflare Workers using the Agents SDK. Load when creating stateful agents, durable workflows, real-time WebSocket apps, scheduled tasks, MCP servers, or chat applications. Covers Agent class, state management, callable RPC, Workflows integration, and React hooks. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.
- ai-elementsAI Elements component library guidance — pre-built React components for AI interfaces built on shadcn/ui. Use when building chat UIs, message displays, tool call rendering, streaming responses, reasoning panels, or any AI-native interface with the AI SDK.
- ai-gatewayVercel AI Gateway expert guidance. Use when configuring model routing, provider failover, cost tracking, or managing multiple AI providers through a unified API.
- ai-generation-persistenceAI generation persistence patterns — unique IDs, addressable URLs, database storage, and cost tracking for every LLM generation
- ai-sdkVercel AI SDK expert guidance. Use when building AI-powered features — chat interfaces, text generation, structured output, tool calling, agents, MCP integration, streaming, embeddings, reranking, image generation, or working with any LLM provider.
- aiq-deploy|
- aiq-research|