SIP URI forwarding into Ocular
How a carrier or a partner (a contact center, the customer's telephony provider) delivers calls from its network to an Ocular voice agent over SIP: the available options, what each side must configure, and what we learned from the first real case, a US partner integrated through its carrier (Bandwidth), which forwards calls to a customer number to Ocular over SIP/TLS.
This page contains no real IP addresses, hostnames, phone numbers or credentials. The concrete values for your integration are provided by Customer Success through the secure channel.
Options for delivering calls to Ocular
| Option | How it works | When it fits | Partner's work | Ocular's work |
|---|---|---|---|---|
| A. SIP URI forwarding (TLS + SRTP) | The partner configures, at its carrier, a forward of the number to an Ocular SIP URI: sip:<DID>@sip.<env>.getocular.ai:5061;transport=tls. The carrier opens the call directly against Ocular's SIP edge. | Partner on a modern carrier (Bandwidth, Twilio, Telnyx…) that requires encryption and DNS resolution. This is the implemented case. | Create the SIP "Location"/destination with the FQDN, port 5061 and TLS; provide the signaling and media IP lists of its carrier; test. | Public domain name, public-CA certificate, dedicated encrypted SIP profile, authorization of the carrier's addresses, number provisioning. |
| B. Plain SIP trunk (UDP/TCP 5060) to an IP | Same as A, without TLS and against the edge's public IP. | Carriers without TLS support, or lab environments. This is how some traditional trunks operate (registration with credentials). | IP:5060 destination, source IPs. | Authorization of the carrier's addresses; no certificate. |
| C. PSTN forwarding to an Ocular DID | The partner forwards its number (call forwarding at its carrier or PBX) to a number Ocular already owns at its own carrier. | When the partner cannot or does not want to touch SIP. Fastest for a trial. | One call forward. | Nothing new if the DID already exists; the caller arrives with the original ANI only if the partner's carrier preserves it. |
| D. Ocular registers as an extension of the partner's PBX | Ocular's edge registers (REGISTER) with the partner's PBX and receives calls like any other extension. | Customers with their own PBX (Asterisk, 3CX, Teams Direct Routing through an SBC). | Create an extension/trunk with credentials. | Registration from Ocular's edge; credentials received through the secure channel. |
Option A is the recommended one for production: end-to-end encryption of signaling and audio, DNS resolution (the IP can change without touching the partner) and isolation in a dedicated SIP profile. The rest of this page focuses on it; B and D reuse the same steps minus TLS.
What to ask the partner before starting
The complete form for the customer is in Inputs we need from the customer. Getting this in writing avoids most of the trial-and-error cycles:
- SIP signaling IPs (source of the INVITEs) and RTP media ranges (source of the audio), per site and per environment. If the carrier has SBC "pairs" or a different range for TLS, ask for both. In the first real case the actual signaling IPs did not match the range the carrier initially "confirmed"; signaling came from a different SBC pair.
- Transport and port they will use: UDP/TCP 5060 or TLS 5061. And whether the "Location" accepts an FQDN or only an IP.
- TLS requirements: version (TLS 1.2 is the norm), cipher suites their SBC offers, whether it sends SNI, whether it requires a public-CA certificate, whether it verifies the name against the FQDN.
- SRTP: mandatory, optional or unsupported, and the profile (
AES_CM_128_HMAC_SHA1_80). - Codecs: G.711 µ-law (PCMU) for North America, G.711 a-law (PCMA) for Latin America and Europe. G.729 only if licensed on both sides.
- Dialed-number format in the Request-URI and
To: E.164 with+, 10- or 11-digit NANP, national. Ocular normalizes at the edge, but you need to know what arrives. - Caller identity (ANI): in
From,P-Asserted-IdentityorRemote-Party-ID. This is what the agent receives ascaller_e164. - DTMF: RFC 2833/4733 (telephone-event) is expected.
- Keepalives: OPTIONS every N seconds, or TLS keepalive. They must be counted separately in monitoring.
- Failover: if they configure a secondary destination, something must exist behind it, or failover masks errors with timeouts.
- Schedule and contact for joint tests, and the evidence they can provide per attempt (Call-ID, SIP code, resolved destination).
What Ocular does with that information
With the carrier sheet, Ocular prepares its side in a few steps and returns, through the secure channel, the destination SIP URI and the addresses your carrier must authorize:
- Network and source control. Only calls from the signaling and media addresses declared by your carrier are accepted. An undeclared source is dropped without a reply, which is why the list must be complete.
- A dedicated SIP profile for your carrier, isolated from the other carriers: a change or a test on your side does not affect other customers, and vice versa.
- A public-CA TLS certificate on an Ocular domain name, compatible with the equipment of common carriers (including those that only negotiate RSA cipher suites or do not send SNI).
- Normalization of the dialed number to international format, according to what your carrier delivers.
- Provisioning of the number with its agent, service hours, after-hours message and overflow policy.
- Monitoring and evidence: signaling capture for your carrier and automatic verification of the profile and certificate status, before and after every change.
- Joint test with your team and your carrier online, following the agreed script and acceptance criteria.
Where a call can stop (for joint troubleshooting)
Ocular network ──► Source control ──► TLS negotiation ──► Number format ──► Number enabled,
(IP/port) (carrier IPs) (certificate) (E.164) hours, agent
Each layer leaves a different trace. When a test fails, the question to your carrier is always the same: which destination it resolved, which transport it used and which reply it received (Call-ID and SIP code). With that, Ocular locates the layer in minutes.
What we learned in the first integration
- Ask your carrier for the real signaling addresses, not just the "official" range. In the first case calls came from a pair of devices that was not on the initial list.
- Confirm transport and port in the destination configured at your carrier: it must point at the domain name and TLS port Ocular provides, not an IP or the unencrypted port, or the negotiation hits the wrong profile.
- Ask which TLS cipher suites your carrier offers. Some devices only negotiate RSA certificates and do not send the server name; Ocular already handles it, but knowing in advance saves a test round.
- A backup destination with nothing behind it turns errors into hard-to-diagnose silence. If you configure one, make sure it exists.
- Agree on the evidence before testing: time, Call-ID and SIP code of each attempt, from both sides.
Checklist
Customer:
- Inputs form sent to Customer Success.
- SIP destination configured at your carrier with the domain name, port and transport Ocular provided.
- Ocular addresses authorized at your carrier, if required.
- Test numbers and test window confirmed.
Ocular:
- Source authorization, SIP profile, certificate and number normalization ready.
- Number provisioned with agent, hours and overflow.
- Signaling capture and automatic verification active.
- Test call completed with shared evidence.