Ready to build better conversations?
Simple to set up. Easy to use. Powerful integrations.
Get free accessReady to build better conversations?
Simple to set up. Easy to use. Powerful integrations.
Get free accessWe’ve opened our AI ecosystem to partners and developers. So, here’s our practical guide to how you can go from an idea to a live integration. One that Aircall's AI agents can call on in real time—using our Developer Portal—to plan, build, and publish.
What you can build with Aircall
Our AI ecosystem gives developers three things to build with:
AI Assist turns every call and message into structured intelligence: transcripts, summaries, sentiment, action items, and QA results, delivered over REST APIs and webhooks.
AI Voice Agents run voice calls end-to-end.
AI Messaging Agents do the same across SMS and WhatsApp.
The connective tissue is the MCP, which lets those agents discover and call your product's capabilities during a live conversation.
The important shift for developers is what an integration now does. Historically, you logged a call or synced a contact after the fact, for a human to read. Now your tool can be something an AI agent reaches for mid-conversation, to look up an order, update a record, or complete a task, so the conversation gets resolved rather than passed along. Connect your product once, and every Aircall AI product can use it.
For example, a retail team could build an Order Concierge; an AI Voice Agent that pulls order details from your commerce platform, processes a refund through a payments provider, and logs the outcome in the CRM, all inside one call. A sales team could build a Lead Qualifier that answers inbound, enriches the record in their CRM, and books a meeting in a scheduling tool before the caller hangs up. Each of those is several connected tools working together in a single conversation, and each tool is something a developer exposed to the agent.
The mental model: connect once, used everywhere
Before writing code, it helps to hold one idea in mind. You aren’t building three separate integrations for three products. You expose your capabilities once, and they become available across AI Voice and Messaging Agents, and AI Assist automatically.
There are two ways in, and most substantial integrations use both. MCP is how you expose actions an agent can take: you describe your tools declaratively with clear schemas and descriptions, stand up an MCP server, and share its URL. Aircall's agents reason over those descriptions and call the right tool at the right moment. REST APIs and webhooks are how you handle classic, event-based work: react to call and contact events in real time, log activity, and sync data into your systems of record.
A quick principle from the planning guide: build event-based integrations. Set up webhooks after user authorization so you react to Aircall events as they happen, rather than polling. And use OAuth for public apps, which gives you secure, scalable authentication and clean installation across many Aircall accounts.
Step 1: Plan your integration
Good integrations start with a plan, not a pull request. Aircall's plan your integration guide lays out four steps.
First, explore what Aircall's API and AI capabilities can do, and pick the ones relevant to your product and users. The feature set ranges from quick, high-impact wins to deeper capabilities. Call logging, insight cards, tag sync, logging AI Voice Agents calls, and triggering AI outbound calls are all low effort and popular. Conversation intelligence such as transcriptions, summaries, playbooks, evaluations, and sentiment, along with SMS and MMS messaging, contact sync, and call management, sits in the higher-value tier.
Second, talk to customers. Identify which Aircall workflows people handle manually today, where the gaps are, and where an agent that can act would save the most time.
Then, write an integration plan: the use cases you want to support, the features you will use, and how it all fits together.
Lastly, share it with us. Once your partner application is approved, submit the plan as a shared document such as a Google Doc, Notion page, or PDF. Expect review feedback within one to five business days, after which you receive developer and OAuth credentials.
Step 2: Build and launch
With credentials in hand, you’ll move into setup and development. The build and launch guide breaks this into clear phases.
Setup comes first. You get a developer account, valid as a trial for four months (and extendable), which becomes permanent once your app reaches its first customers. You also receive two OAuth applications, one for development and one for production, with support for up to five redirect URIs each. One practical note that saves time: localhost URIs aren’t supported, so use a tunneling service such as ngrok while you develop.
A few OAuth specifics are worth knowing up front. An admin install grants a single account-level access token for the whole account. You match Aircall users to your own by email using the List all Users endpoint. And admins choose which phone numbers your integration can access during the OAuth flow.
Then you build. It helps to know the two shapes an app can take:
A public app is designed for many Aircall customers, uses OAuth, and goes through review.
A private app is a single-account integration that can use API keys and needs no review.
For the AI ecosystem, the MCP work is deliberately light: expose your actions as tools, keep schemas and descriptions clean, and let the agents do the rest. There’s no brittle glue to maintain, because the integration is declarative. When a workflow changes, you update the schema, not a pile of code.
Aircall asks that integrations reach completion within about six months of receiving credentials, with a short progress update at least once every couple of months. That cadence keeps momentum and keeps your credentials active.
Step 3: Publish your app
Publishing is how your integration becomes discoverable to Aircall's customers. Our how to publish an app guide sets out the requirements and the assets you will need.
To be eligible to publish a public app, you need a finished, live product with a company website, and an existing customer base using it. Meeting the bar doesn’t guarantee approval, since Aircall keeps a discretionary review, but it’s the baseline. You apply through the technology partner form, and applications are reviewed within about five business days.
When you’re ready to list, prepare four assets: a 5-10 minute demo that walks through the integration; at least two help center articles covering installation and your main features; a logo in SVG format using Aircall's template; and 3-5 screenshots that show the value of the integration. You submit these through the official form, then demo and asset review takes one to five business days, and publication follows in three to five.
There’s one more milestone worth planning for: new apps live in the Aircall Dashboard until they reach 10 customers. Once you cross that mark, you email Aircall to get listed in the public marketplace, and your developer account becomes permanent.
Why it’s worth building here
The economics favor building early. One MCP server works across every Aircall AI product at once, so you aren’t maintaining three integrations. Because MCP is becoming the interoperability standard, the same server you build for Aircall works with any other MCP-speaking platform, so your effort isn’t locked to one vendor. Your tool stops being a passive integration and becomes infrastructure that other companies' AI agents depend on mid-conversation.
Publishing adds the distribution. A marketplace listing brings discoverability and the credibility of a reviewed, maintained integration, plus access to co-marketing, technical guidance, and early API features. Partners who integrate now are visible to Aircall's customers from day one, while the ecosystem is still small enough to stand out in!
Start building
Everything begins in the Developer Portal, and the fastest path is to plan while your partner application is under review, so you can move quickly once approved.
Apply as a technology partner, map your first use case to the capabilities above, and ship an integration that doesn’t just watch conversations, but helps resolve them.
Published on September 2, 2026.
