Skip to content

Class 4 Softswitch

Document Metadata
Category: Infrastructure & Connectivity / SIP Devices
Audience: Network Engineers, Administrators, Support Team
Difficulty: Intermediate to Advanced
Time Required: Approximately 30–45 minutes
Prerequisites: Active ConnexCS account with permissions to manage SIP devices (DIDs, gateways, SIP endpoints) and access to Device Configuration modules.
Related Topics: SIP Device
Next Steps: After configuring your SIP devices, perform end-to-end call tests (inbound/outbound), monitor device registration and usage in the Reporting module, and schedule regular review of firmware and configuration settings for security compliance.

Overview

ConnexCS is a cloud-based Class 4 Softswitch designed to route and manage VoIP calls between different telephony networks.

A Class 4 Softswitch operates at the network level. It receives a call, determines the appropriate destination based on configured routing rules, and forwards the call to the selected carrier or downstream system.

ConnexCS does not provide the end-user call endpoint. Instead, it acts as an intermediary between customers, carriers, and other VoIP systems, providing the routing and switching layer that connects different networks.

ConnexCS provides capabilities such as:

  • Call routing based on configured routing strategies
  • Carrier selection based on destination, rates, quality, and other criteria
  • Number and CLI manipulation
  • Call authorization and validation
  • Billing and usage management
  • Traffic and call monitoring
  • Interconnection between multiple VoIP networks

For example, when a customer sends a call through ConnexCS, the platform evaluates the call against the configured routing rules and determines which carrier should receive it. ConnexCS then routes the call to the selected carrier.

Customer → ConnexCS Class 4 Softswitch → Carrier → Destination

Class 4 vs. Class 5 Softswitch

A Class 4 Softswitch is primarily designed for handling and routing traffic between telecommunications networks. It focuses on high-volume call routing, switching, interconnection, and traffic management.

A Class 5 Softswitch, on the other hand, is designed for end-user services and provides features such as voicemail, call forwarding, announcements, and other subscriber-facing functionality.

ConnexCS is a Class 4 Softswitch and can connect to Class 5 Softswitches, dialers, and other VoIP applications that require a routing platform.

Cloud-Based Platform

ConnexCS is delivered as a Platform as a Service (PaaS). This allows customers to use a globally distributed and scalable Class 4 switching infrastructure without managing the underlying switching hardware or infrastructure.

Customers can connect their VoIP systems, carriers, and applications to ConnexCS and use the platform to manage and route their voice traffic across multiple networks.

Class 4 Softswitch

Routing Engine

The Class 4 platform routes calls using one of the following strategies, configurable per route or per prefix:

  • Least Cost Routing (LCR) — selects the lowest-cost healthy carrier for a given destination.
  • Round Robin — distributes calls sequentially across a defined carrier set.
  • Weighted Random — distributes calls probabilistically according to assigned weight values per carrier.
  • Tiered Distribution — applies a priority order across carrier tiers, falling through to lower tiers on failure or exhaustion of higher tiers.

Custom routing logic beyond these built-in strategies can be implemented using ScriptForge, a JavaScript-based scripting environment that allows routing decisions (LCR, round robin, weighted, tiered, or fully custom logic) to be defined programmatically per route.

See: Routing Strategy, Routing Overview

ASR+ (Answer-Seizure Ratio Prediction)

ASR+ is a connectivity-prediction mechanism that evaluates cached historical call data to estimate the likelihood that a given number will connect successfully, prior to the call leaving the switch.

  • Numbers with a low predicted connection probability can be filtered before dialing, reducing carrier setup attempts on non-connecting numbers.
  • The mechanism operates on a rolling cache of prior call outcomes per number/prefix rather than real-time carrier state.

See: Ingress Routing / ASR+ details

Rate Card Engine

Customer rate cards are derived from carrier (provider) rate cards through configurable margin rules, rather than being built or maintained as separate flat files.

Supported capabilities:

  • NPA-NXX and LRN routing, including unlimited LRN dips at no additional per-dip cost.

  • Multi-currency support with configurable rounding behavior.

  • Full revision history retained for every rate card, including carrier and customer cards.

  • Margin rules applied at compile time when generating a customer card from one or more carrier cards.

See: Rate Card Overview, Provider Rate Card, Customer Rate Card

CPS Buffering

The platform can buffer inbound call-per-second (CPS) bursts that exceed a configured carrier's CPS limit, smoothing traffic spikes rather than rejecting calls that arrive above the limit.

  • Buffering capacity is rated up to 5x the configured carrier CPS limit.
  • CPS and channel caps are configurable independently per customer, per carrier, and per route.

See: Channel Limitation Settings, Capacity Limiting

Ingress Architecture — AnyEdge

A single SIP ingress point represents a single point of failure. AnyEdge provides an Anycast-based, globally distributed SIP entry layer in front of the Class 4 switch.

  • Traffic is routed to the nearest healthy edge node via Anycast.
  • Each node supports 10,000+ CPS capacity and 10 Gbps DDoS mitigation.
  • If a node becomes unhealthy, traffic is redirected to the next-nearest healthy node without manual intervention.

See: AnyEdge

MCP Server (AI Agent Integration)

The ConnexCS MCP (Model Context Protocol) Server exposes platform diagnostics, call data, and reporting functions to AI agents (e.g., Claude) over a JWT-secured connection.

Supported operations include (non-exhaustive):

  • searchCallLogs — query historical call log data.

  • getSipTrace — retrieve SIP signaling traces for a given call.

  • investigateCall — run a diagnostic investigation across routing, carrier response codes, and failover events for a given call or time window.

The MCP Server sits between the AI agent and the platform's internal APIs, and is accessed via ScriptForge tool functions or the CX MCP App layer.

See: MCP Server

Compliance & Fraud Controls

Regional compliance and fraud mitigation are enforced at the platform level:

  • STIR/SHAKEN attestation and certificate management (USA), including LRN routing/free dips and FCC-oriented compliance tooling.
  • TPS scrubbing (UK) and other region-specific regulatory controls.
  • Fraud profiles and capped rate limits to constrain exposure from anomalous traffic patterns.
  • DDoS protection enforced at the ingress layer (see AnyEdge, above).
  • Granular permissions covering 80+ definable functional areas, with support for hardware-based two-factor authentication (2FA).

See: Security, STIR/SHAKEN Cert, Fraud Profile

Observability

  • Always-on SIP capture allows any call to be reconstructed and rendered as a signaling ladder diagram after the fact, including calls from prior weeks, subject to retention limits.
  • Real-time CDRs are generated per call.
  • Circuit tests measure MOS (Mean Opinion Score), PDD (Post-Dial Delay), and DTMF signaling accuracy on demand.
  • Diagnostic data captured through this layer is queryable by the MCP Server (see above).

See: Logging, SIP Traces, Pings and Messages