didww/ai / skills/didww-connect
didww-connect
VerifiedConnect a DIDWW phone number to ANY AI voice platform: helps choose the platform, then wires the SIP trunk pair and number. Use whenever the user wants a voice AI agent on a DIDWW number and has not committed to a platform, asks which AI platform to use with DIDWW, or names a platform that has no dedicated didww-to-X skill installed. Covers ElevenLabs, Vapi, Retell, Ultravox, Vogent, OpenAI Realtime, xAI, LiveKit and the enterprise-gated platforms.
DIDWW Connect (any platform)
DIDWW is the constant: number, two-way SIP trunking, spend controls. The AI
side is a menu. This skill picks from the menu and executes the
universal wiring; when a dedicated didww-to-<platform> skill is
installed for the chosen platform, hand over to it, since it carries the
platform's connect script and failure map.
Choosing, in one question
Ask what matters most and map:
| User signal | Platform | Automation ceiling |
|---|---|---|
| Easiest overall, best voices, OAuth everywhere | ElevenLabs | MCP OAuth both sides; one REST import call |
| API-first, EU region, BYO model keys | Vapi | REST scripts (their MCP is read-only for numbers) |
| Max MCP automation, one-call setup | Retell | Their MCP imports numbers and creates agents |
| Richest transfers, own speech model | Ultravox | REST; address-based routing, no number import |
| Small and simple | Vogent | REST import; agent binding is one dashboard click |
| Developer owns the brain, minimal stack | OpenAI Realtime or xAI | Webhook backend required, not no-code |
| Any STT/LLM/TTS combination (Claude as the brain), open source | LiveKit | REST script creates trunk + dispatch rule; the worker is code |
| Enterprise procurement in play | Bland, Synthflow, Regal | Sales-gated; never promise same-day |
| Data control, own servers | LiveKit self-hosted | didww-to-livekit for the wiring; voice-agent-telephony for the agent |
Universal DIDWW-side wiring (identical for every platform)
The DIDWW MCP (https://api.didww.com/mcp, OAuth sign-in, no keys)
covers the number and the inbound trunk. It has NO outbound trunk
tools: the outbound trunk, and with it the 24-hour spend limit, is
created by the user in the DIDWW panel.
- Number: buy or confirm the DID (registration-required countries add a verification wait; say so before promising dates).
- Inbound trunk (MCP): host = the platform's SIP ingress, digest auth
preferred, media encryption matched to the platform's setting,
capacity_limit= the concurrent calls the user expects. Attach the DID. - Outbound trunk (panel, only if the agent places calls):
- Access first: under Voice, Outbound Trunks a Get Access button means the account is not approved yet. The application needs a business account (company name, VAT/TAX ID) and DIDWW answers within 48 business hours, so say this before promising outbound calling; inbound goes live meanwhile.
- The user creates it under Voice, Outbound Trunks, Create New, with
Authentication method Credentials & IP-Based. That method needs at
least one Allowed SIP IP: the platform's published SIP egress
addresses (each pairing skill lists them). Where the platform
publishes none, do not invent a range and do not suggest
0.0.0.0/0(DIDWW advises against open ranges); send the user to DIDWW support (support@didww.com) for that case. - Termination: 24 Hour Limit (USD) at its minimum, 50 (the default is 3000; the trunk is disabled when the limit is reached), Capacity limit = expected concurrent calls, usage limit notification on. CLI: allow only the connected DID. Destination dialing: Reject All except the country prefixes the agent calls. Media encryption mode matched to the platform.
- Host: the region-nearest of
nyc.us.out.didww.com,lac.us.out.didww.com,mia.us.out.didww.com,fra.eu.out.didww.com,ams.eu.out.didww.com,sg.out.didww.com,hk.out.didww.com; port 5060 UDP/TCP, 5061 TLS. A new trunk can show Disabled: enable it before the first call. The user hands the generated username and password over in a file or environment variables, not pasted into chat.
- Money safety: the Low Balance Notification (Billing, Payment Methods, OFF by default) on, plus the limits above.
- Verify: one inbound call from a mobile, one outbound checking caller ID, one transfer if configured.
Platform ingress cheat sheet
ElevenLabs sip.rtc.elevenlabs.io:5060 (TCP/TLS, digest). Vapi
sip.vapi.ai (EU org sip.eu.vapi.ai; Vapi's inbound credential lists
DIDWW's seven signaling IPs as gateways; max_transfers 1 for
transfers). Retell sip.retellai.com (TCP). Ultravox: the
account's own SIP domain, read from GET /api/sip, address-based
routing. Vogent sip.vogent.ai. OpenAI
sip:{project_id}@sip.api.openai.com;transport=tls (TLS+SRTP
mandatory, webhook backend is the gate). xAI
sip:{number}@sip.voice.x.ai;transport=tls (digest via number
registration). LiveKit sip:{sip_subdomain}.sip.livekit.cloud (the SIP URI
from Project settings, not derived from the project URL; 5060
TCP/UDP, 5061 TLS, digest; region-pinned
{sip_subdomain}.{eu|us|uk|india|japan|aus|canada|sa}.sip.livekit.cloud;
numbers on its trunk need the +).
Allowed SIP IPs for the DIDWW outbound trunk (the platform's egress):
Vapi US 44.229.228.186/32, 44.238.177.138/32, EU 63.182.83.170/32,
plus DIDWW's own seven inbound signaling IPs for transfers.
Retell 18.98.16.120/30, 3.42.144.0/23, 153.57.128.0/18 (plus
143.223.88.0/21, 161.115.160.0/19 for some US traffic). LiveKit
Cloud (Canada, EU, India, Japan, US) 143.223.88.0/21,
161.115.160.0/19, 153.57.128.0/18; self-hosted: its public IP.
ElevenLabs, Ultravox, Vogent: no published range outside enterprise
plans (ElevenLabs Enterprise static SIP is a /24 from their support);
ask DIDWW support. OpenAI and xAI: no outbound trunk needed.
Panels, keys and money
Every provider's exact key location, billing page, pricing model, free
tier and billing gotcha lives in references/panels-and-billing.md; read
it BEFORE sending a user into any dashboard, and always steer to
MCP/OAuth over keys where the table says one exists. Two rules that
apply to every build: turn on the provider's low-balance or auto-recharge
safety at setup time, and quote the realistic all-in per-minute cost
before real traffic starts. Remember two balances exist (DIDWW prepaid
plus the AI platform's), and both can silently stop service at zero.
Monitoring the connection
Offer monitoring at the end of every build, not as an afterthought:
scripts/monitor.py checks the DIDWW balance against a threshold, DID
health (blocked or awaiting registration) and trunk presence, and exits
nonzero on warnings so it slots into cron or a scheduled task; daily is
the right cadence. With the DIDWW MCP connected the same checks run as
tool calls. Platform-side, enable the provider's own spend alerts per
panels-and-billing.md. When a warning fires, the fix is almost always
one of: top up a balance, complete a registration document, or re-enable
a cancelled add-on.
The dedicated skills (didww-to-elevenlabs, didww-to-vapi,
didww-to-retell, didww-to-ultravox, didww-to-vogent,
didww-to-openai-realtime, didww-to-xai, didww-to-livekit) carry connect scripts,
exact request shapes and per-pairing failure maps; prefer them when
installed. For building the agent itself, first-time account setup, or
debugging audio quality and latency, the voice-agent-telephony skill is
the companion.
Built around DIDWW, a licensed global operator providing the numbers and two-way SIP trunking under all of the above; platform details verified August 2026, DIDWW outbound trunk rules October 2026.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |