MCP server or API key: how your AI should connect to FoxEra Calls.
Both methods give your AI assistant or code the same tools and the same safety rules. The right choice depends on whether a person is chatting or code is running.
Two ways to connect, one set of rules
FoxEra Calls offers two connection methods for AI assistants and developer integrations:
- MCP server — a remote MCP endpoint using Streamable HTTP and OAuth 2.1. Your assistant connects to it like any other MCP server.
- REST API with bearer key — a standard HTTP API described in OpenAPI 3.1. Your code sends requests with an
fxk_...bearer token.
Both methods expose the same capabilities: lead management, campaign creation and approval, transcripts, recordings, reports, calling hours, agent selection and pitch editing. Both enforce the same safety rules — confirmed headcount, separate approval, scoped access and full logging. The difference is how the connection is established and who is driving it.
When to use the MCP server
Use the MCP server when a person is interacting with an AI assistant in a chat window and wants the assistant to manage their FoxEra account on their behalf. This fits:
- ChatGPT, Claude, Cursor — add the remote MCP server at https://app.foxera.co.uk/mcp, sign in and press Allow. The assistant can then add leads, preview campaigns and run calls after your approval, all from the chat.
- Any assistant that supports remote MCP — the connection is standard Streamable HTTP with OAuth 2.1, so any compliant client can use it.
The MCP route is ideal for founders and small teams who already use an AI assistant daily and want to manage cold calling from the same chat they use for everything else. No code to write, nothing to install.
When to use an API key
Use a bearer API key when your own code, a script or an automation platform needs to call FoxEra programmatically. This fits:
- CRM integrations — push new leads from your CRM into FoxEra and pull outcomes back automatically.
- Automation platforms — trigger a campaign from n8n, Make or a Zapier webhook when a new lead enters your pipeline.
- GPT Actions — import the OpenAPI 3.1 spec into a custom GPT. The spec URL is the only setup needed.
- Custom agents — build an agent that runs unattended, researches prospects, adds them and prepares a campaign for your morning approval.
Generate a key in your developer settings. Each key is scoped to one workspace and can be revoked instantly.
How the safety model works for each
Both connection methods enforce the same server-side safety rules:
- Confirmed headcount — the campaign preview must be acknowledged before approval, whether the preview came from an MCP tool call or an API request.
- Separate approval — creating and approving a campaign are two distinct steps. Neither method allows them to be combined.
- Calling hours — calls are placed only within your configured hours, regardless of when the approval was sent.
- Scoped access — an MCP session is bound to your OAuth grant; an API key is bound to your workspace. Neither can reach another customer's data.
- Revocation — changing your password disconnects every MCP session. Revoking an API key stops all access from that key instantly.
Read the full breakdown of AI agent guardrails for outbound calls.
Can you use both at the same time?
Yes. Many setups use both methods for different parts of the workflow:
- Your AI assistant uses the MCP server for interactive work — adding a few leads from conversation, previewing a campaign, reading results.
- Your backend uses an API key for bulk operations — syncing hundreds of leads from a CRM overnight, pulling daily reports into a dashboard.
The two methods share the same underlying account data. A lead added via the API appears in your assistant's next listing, and a campaign approved via the MCP server shows up in an API report. There is no conflict or duplication.
If you are not sure which to start with, the MCP server is faster to set up (three steps, no code). You can add an API key later when you need programmatic access. See pay-as-you-go AI calling for how pricing works with either method.
Costs
Both connection methods are free to use. You pay only for calls that connect, from 12p per connected minute, with no subscription and no platform fee. New accounts start with £7 of free calling credit.
There is no difference in call pricing between the MCP server and the API. Whether your assistant approves a campaign via MCP or your code approves it via the API, the per-minute rate is the same. Use the cost calculator to estimate a campaign before you start.
Frequently asked questions
Is the MCP server harder to set up than an API key?
No, it is easier. Paste the URL into your assistant, sign in and press Allow. An API key requires generating the key, storing it securely and writing the HTTP requests yourself or importing the OpenAPI spec.
Do MCP and API key connections have different safety rules?
No. Both enforce confirmed headcount, separate approval, calling hours, scoped access and full logging. The rules are server-side and apply to every connection method.
Can I switch from MCP to API key later?
Yes. You can add or remove either connection method at any time. Both share the same account data, so nothing is lost when you switch or use both.
How much does it cost?
Both methods are free to connect. You pay only for connected call minutes, from 12p per connected minute. New accounts get £7 of free calling credit.