Aircall is inviting a small group of customers into a private beta of something its developer documentation has hinted at for weeks: an Aircall-hosted MCP server that connects AI assistants like Claude and ChatGPT directly to your Aircall call data. Instead of exporting reports or writing integration code, an authorized admin can ask questions in plain language — "search my calls from yesterday," "what action items came out of last week's demos?" — and get answers pulled from live call records, transcripts, summaries, and sentiment data.
The invitation reached customers by email in early October 2026. Aircall has not announced the beta publicly, and the details below come from that invitation, cross-checked against Aircall's public developer documentation and its September 2026 AI ecosystem announcement. If you did not receive an invitation, treat this as a preview of what is coming and a checklist for getting ready.
Quick Answer
Aircall MCP is a private beta that connects an AI assistant — Claude Desktop, Claude Code, or ChatGPT/Codex — directly to your Aircall call data through the Model Context Protocol (MCP). As of October 2026 it is read-only, limited to calls (no SMS, MMS, or WhatsApp), and available only to Aircall admin users who have the Aircall AI add-on. It matters for sales and service leaders who want conversation intelligence queryable from the AI tools their teams already use, without a custom integration build. Vantage Point helps teams evaluate AI-and-telephony changes like this against their CRM, data governance, and compliance requirements before they connect anything.
TL;DR
- What it is: Aircall's own MCP server (private beta) that lets Claude or ChatGPT query your Aircall calls, transcripts, summaries, and sentiment in plain language.
- Current limits: Read-only, calls only, Aircall admins only, and it requires the Aircall AI add-on for the conversation-intelligence tools.
- Connection: One-time OAuth, like connecting any other app — no integration code. Aircall provides the server URL to beta participants.
- Decision point: Which admins get access, and which call data should be queryable by an AI assistant at all — recordings and transcripts can contain sensitive client information.
- How Vantage Point helps: We implement Aircall alongside Salesforce and HubSpot and can help you set access, governance, and CRM integration policy before you join a beta like this.
Status as of October 2026: everything in this post reflects a private beta invitation and Aircall's public documentation as of October 6, 2026. Beta scope, eligibility, and limits can change at general availability. Verify details against your own Aircall invitation or account team.
What Changed: Aircall Now Has Its Own MCP Server (in Private Beta)
MCP — the Model Context Protocol — is an open standard that lets AI assistants discover and use external tools and data sources over a consistent interface. We covered the fundamentals in our plain-English explainer on MCP. The short version: instead of building a custom API integration for every question you want an AI to answer, you connect the assistant to an MCP server once, and it can call the tools that server exposes.
Aircall's developer FAQ, updated in early September 2026, stated that Aircall did not yet offer its own MCP server but told readers to "stay tuned." The private beta invitation customers received in early October is that news arriving: Aircall is running an MCP beta with a small group of invited customers. The server gives an AI assistant read access to your call data — search, transcripts, summaries, action items, and sentiment — through natural-language prompts.
Example prompts from Aircall's invitation:
- "Search my calls from yesterday."
- "Pull the transcript and summary of this call."
- "What action items came out of last week's demos?"
- "Show me calls with negative sentiment."
For teams already working in Claude or ChatGPT, that turns conversation intelligence into something you query in the same window where you draft follow-ups, prepare for renewals, or summarize a week of pipeline activity — no export, no dashboard, no code.
Two Directions of Aircall's AI Ecosystem — Don't Confuse Them
Aircall's September 2026 announcement, "Aircall opens its AI to the ecosystem", described the company opening its platform to AI in both directions across a base of 23,000 businesses and 250+ integrations. The beta is the second of two directions, and it helps to keep them straight:
| Direction | What it does | Status (as of Oct 2026) |
|---|---|---|
| Partners expose tools to Aircall AI (AI Actions / Partner MCP Registration) | External partners register their own MCP servers so Aircall's AI agents can act inside other tools — Shopify, HubSpot, Zendesk — during a call. | Live. Public "Register a Public MCP" documentation and MCP FAQs in the Aircall Developer Portal. |
| Aircall exposes call data to your AI assistant (Aircall MCP) | Claude or ChatGPT connects to Aircall's own MCP server and queries your calls, transcripts, summaries, and sentiment in plain language. | Private beta with invited customers. Documented for participants in a support article titled "Using Aircall MCP with AI assistants," which is not yet publicly listed. |
The first direction is about Aircall's agents reaching outward and taking action in your stack — we covered the CRM side of that in our guide to Aircall AI Actions. The new beta is the read side of the ecosystem opening: your AI assistant reaching inward to Aircall's conversation data.
Who Is Affected — and Who Can Join
This beta is narrow by design. Per Aircall's invitation, as of October 2026:
- Invitation-only. Aircall invited a small group of customers. If you were not invited, ask your Aircall account team about access — there is no public sign-up page.
- Admin users only. You must be an Aircall admin to connect. Aircall says support for other user types is coming soon, which suggests broader roles at or after general availability.
- API key permissions required. The connecting user needs to be able to create API keys (Settings → Integrations → API keys in the Aircall dashboard).
- Aircall AI add-on required. The conversation-intelligence tools — transcripts, summaries, action items, sentiment — depend on the Aircall AI add-on. Without it, expect a much thinner set of capabilities.
The teams this matters most for are sales and service leaders whose reps live on the phone and whose managers already use Claude or ChatGPT for analysis and drafting. If your calls log to Salesforce or HubSpot today, this does not replace that integration — it adds a question-answering layer on top of the same call data you already review in Aircall's conversation-intelligence dashboards.
What the Beta Can and Cannot Do
| Capability | In the beta? |
|---|---|
| Search and retrieve calls by natural-language query | Yes |
| Pull transcripts, summaries, and action items | Yes (with the Aircall AI add-on) |
| Filter calls by sentiment (e.g., negative sentiment) | Yes (with the Aircall AI add-on) |
| Make calls, send messages, tag calls, or change anything in the account | No — strictly read-only |
| SMS, MMS, or WhatsApp conversations | No — calls only |
| Access for non-admin Aircall users | Not yet — "coming soon" per Aircall |
The read-only scope is the right call for a beta and worth appreciating rather than lamenting: an AI assistant that can query call recordings and transcripts can also surface sensitive information to whoever holds the connection. Keeping write actions out of scope removes an entire category of risk while Aircall and its customers learn how these tools behave.
How the Connection Works
For beta participants, setup is deliberately simple:
- Confirm you are an Aircall admin who can create API keys (Settings → Integrations → API keys) and that your account has the Aircall AI add-on.
- In your AI assistant — Claude Desktop, Claude Code, or ChatGPT/Codex — add a new MCP server connection. Aircall provides the server URL to beta participants in the invitation and participant documentation.
- Complete the one-time OAuth authorization, exactly like connecting any other app to Aircall.
- Start asking questions against your call data.
There is no integration code to write and nothing to deploy. The practical work is not technical — it is deciding who should hold the connection and what they should be able to ask.
What to Do Now: A Five-Step Readiness Checklist
- Verify your licensing. Confirm whether your Aircall plan includes the AI add-on. The most useful beta features (summaries, action items, sentiment) depend on it, and it is a separate line item.
- Decide which admins get access. Because the beta is admin-only, every connection is a privileged one. Limit it to admins with a real need, and keep a list of who connected which assistant and when.
- Define your data policy before connecting. Call recordings and transcripts frequently contain sensitive client information — and in regulated industries, potentially records subject to retention, supervision, or privacy obligations. Decide which call data should be queryable by an AI assistant at all, whether queries happen inside or outside your approved AI tooling, and how that fits your existing compliance and security posture.
- Test against real workflows. Good beta tests are specific: weekly pipeline-call reviews, QA sampling for support calls, action-item extraction after demos, sentiment spot-checks on at-risk accounts. Compare the assistant's answers against Aircall's own dashboards before trusting them in a workflow.
- Give feedback during the beta. Private betas are short and shaped by participants. If you want non-admin roles, messaging channels, or specific CRM tie-ins, say so while Aircall is still deciding what general availability looks like.
Timeline and What to Watch
- September 1, 2026: Aircall announced its open AI ecosystem — AI Actions over MCP, the Developer Portal, and Partner MCP Registration — and its developer FAQ said an Aircall-hosted MCP server was not yet available but to "stay tuned."
- Early October 2026: Aircall began inviting a small group of customers to the private MCP beta.
- Next: Aircall has not announced a general-availability date. Watch for three changes that would materially expand the release: access beyond admin users (flagged as "coming soon"), coverage of SMS/WhatsApp conversations, and any move beyond read-only actions. Each one changes both the value and the governance calculus.
How Vantage Point Helps
Vantage Point is an Aircall partner that implements cloud telephony alongside Salesforce and HubSpot, so we see these changes from both sides: the phone system and the CRM it feeds. If you are weighing the Aircall MCP beta, we can help you decide which admins and workflows belong in scope, align call-data access with your compliance and security requirements, make sure the underlying Aircall–CRM integration is logging cleanly so AI answers rest on accurate data, and fold telephony AI into a broader CRM and AI roadmap rather than treating it as a one-off experiment.
Frequently Asked Questions
What is the Aircall MCP beta?
It is a private beta of Aircall's own MCP server, which connects AI assistants such as Claude and ChatGPT directly to your Aircall call data. Once connected, you can ask plain-language questions — searching calls, pulling transcripts and summaries, extracting action items, or filtering by sentiment — without writing integration code.
Who can join the Aircall MCP beta?
Aircall invited a small group of customers by email in early October 2026, and there is no public sign-up. Within invited accounts, only Aircall admin users can connect, and the admin must be able to create API keys. If you were not invited, ask your Aircall account team about access.
Is Aircall MCP read-only?
Yes, as of October 2026 the beta is strictly read-only. The assistant can retrieve call data but cannot make calls, send messages, tag calls, or change anything in your Aircall account. Aircall has not said whether write actions are planned for general availability.
Does Aircall MCP include text messages or WhatsApp?
No. The beta covers calls only — SMS, MMS, and WhatsApp conversations are not included. Teams that rely heavily on messaging channels should treat this as a call-intelligence tool for now.
Do I need the Aircall AI add-on to use the MCP beta?
You need the Aircall AI add-on for the conversation-intelligence tools — transcripts, summaries, action items, and sentiment — that make the beta most useful. Basic call retrieval is possible without it, but the add-on is effectively a requirement for the workflows Aircall highlights.
Which AI assistants work with Aircall MCP?
Per Aircall's invitation, the beta works with Claude Desktop, Claude Code, and ChatGPT/Codex. Connection is a one-time OAuth authorization, and Aircall provides the server URL to beta participants.
What should we do about call-recording privacy before connecting?
Decide your policy before anyone connects: which admins get access, which call data should be queryable by an AI assistant, and whether transcripts containing regulated client information may leave your approved tooling. Vantage Point can help align that policy with your compliance, security, and CRM governance requirements.
Evaluating Aircall's AI Features for Your Team?
Vantage Point implements Aircall, Salesforce, and HubSpot as one connected stack — telephony, CRM, and the AI layer on top. If you want help scoping the MCP beta, tightening call-data governance, or cleaning up the integration underneath it, start a conversation with our team.
