Build your own AI cold caller with Twilio — or just connect one that already works.
If you are a developer weighing a DIY voice stack against a ready-made service, here is what building actually requires, what it costs beyond the API bills, and what you get instead by connecting FoxEra's MCP server or REST API.
What building an AI cold caller actually requires
A working AI cold calling system is not one API call. It is several services wired together, each with its own pricing, rate limits, failure modes and compliance surface. Here is the minimum stack:
- Telephony provider — Twilio, Vonage or similar. You need outbound calling, UK number provisioning, call recording and webhook handling for call events.
- Speech-to-text — real-time transcription of the prospect's side of the call, fast enough to keep the conversation natural.
- Language model — OpenAI, Anthropic or another provider to generate the agent's responses based on the transcript and your pitch.
- Text-to-speech — converting the model's output back to audio in real time, with a voice that sounds natural on the phone.
- Agent logic — managing turn-taking, handling interruptions, knowing when to stop talking and when to wrap up the call.
- Recording and transcript storage — saving the audio and text for review, compliance and follow-up.
- UK calling-hour rules — ensuring calls are only placed during permitted hours and that TPS/CTPS screening is handled.
Each of these is a separate integration, a separate bill, and a separate thing that can break at 2 a.m. on a Tuesday.
The costs you do not see in the API pricing page
The API bills are the easy part. Twilio charges per minute, OpenAI charges per token, the TTS provider charges per character. You can estimate those before you start.
The harder costs are:
- Development time — wiring real-time audio streaming between telephony, STT, the model and TTS with acceptable latency is not a weekend project. Expect weeks of work before a single call sounds right.
- Ongoing maintenance — API changes, rate limit adjustments, model deprecations. OpenAI has deprecated models mid-project more than once. Each change is your problem.
- Compliance — UK cold calling rules require TPS screening, calling-hour limits and disclosure requirements. Building these into your stack correctly is your legal responsibility.
- Monitoring — knowing when a call fails, when latency spikes, when the model hallucinates something your business should not be saying. You need alerting, logging and a way to review calls after the fact.
None of these appear on an API pricing page, and all of them are reasons finished products exist.
What connecting an existing service gives you
FoxEra Calls handles the entire voice stack — telephony, speech, agent logic, recording, transcription and UK compliance — as a single service. You interact with it through an MCP server or REST API, which means your code (or your AI assistant) can:
- Add leads to a folder
- Choose a calling agent (Cass, Alex or Andrea)
- Set or edit the calling script
- Preview the campaign cost and headcount
- Start the campaign after approval
- Retrieve call transcripts and outcomes
You write the integration logic — the part that is specific to your business. FoxEra handles everything below the API line: the telephony, the real-time audio, the agent behaviour, the recording, and the script execution.
Comparing the two approaches side by side
The trade-off is control versus speed. Building gives you full control over every layer of the stack. Connecting gives you a working system today.
- Time to first call — building: weeks to months. Connecting: the same day you add your leads and approve a campaign.
- Cost per minute — building: Twilio plus STT plus LLM plus TTS, summed across providers. Connecting: from 12p per connected minute, all included, no subscription.
- Maintenance — building: every API change is yours to handle. Connecting: FoxEra handles the stack; you maintain your integration logic only.
- Compliance — building: your responsibility to implement correctly. Connecting: calling-hour rules and TPS screening guidance are handled by the platform.
- Flexibility — building: you can change anything. Connecting: you work within the API surface, which covers most B2B cold calling workflows but not every edge case.
When building your own stack makes sense
There are cases where building is the right call. If you need a voice agent that does something FoxEra does not support — inbound customer service, appointment scheduling with calendar integration, or a conversational flow that goes beyond cold calling — then a custom stack may be necessary.
If you are building a product rather than using a tool — say, a vertical SaaS that includes outbound calling as a feature — owning the stack gives you the flexibility to differentiate.
But if your goal is to phone a list of UK businesses, introduce your company, and log the results so you can follow up the warm ones, building from scratch is solving a problem that is already solved. The development time is better spent on the parts of your business that are genuinely unique.
Getting started with the FoxEra API or MCP server
The developer guide covers both the MCP server (for AI assistants like ChatGPT, Claude and Cursor) and the REST API (for direct integration from any language). Connect at https://app.foxera.co.uk/mcp for MCP, or use the REST endpoints documented in the guide.
New accounts start with £7 of free calling credit, which is enough to test the API end to end — add leads, preview a campaign, approve it, and pull the transcripts. Calls cost from 12p per connected minute after that.
If you want to see how the API compares to manual cold calling or a traditional dialler, those guides cover the operational comparison.
Frequently asked questions
How much does it cost to build an AI cold caller from scratch?
The API costs (Twilio, OpenAI, TTS) vary by volume, but the real cost is development time — typically weeks of work before the first call sounds right, plus ongoing maintenance. FoxEra costs from 12p per connected minute with no development required.
Can I connect FoxEra to my own application via API?
Yes. FoxEra offers both an MCP server (for AI assistants) and a REST API (for direct integration). The developer guide at foxera.co.uk/developers covers both options.
Does FoxEra handle UK cold calling compliance?
FoxEra provides calling-hour rules and TPS screening guidance as part of the platform. You are still responsible for the quality of your lead list and your calling script. See the UK cold calling rules guide.
Can I switch from my own build to FoxEra later?
Yes. The API accepts leads in standard formats, so migrating a contact list is straightforward. You can run both systems in parallel during a transition.