WhatsApp makes the leap to third-party AI agents: here's how the new integration will work

  • WhatsApp is testing in its Android beta version the connection of up to five AI agents or third-party bots per account, with dedicated and independent chats.
  • Chats with these agents will not have end-to-end encryption, a significant change from regular conversations.
  • The feature responds to European regulatory pressure and paves the way for Meta's strategy with conversational agents.
  • Meta has not confirmed a general rollout date, but the beta is already active for a small group of users.

WhatsApp AI agents

WhatsApp is taking firm steps to become more than just a messaging app. The company has begun testing in its Android beta version the ability to connect to third-party AI agents and chat with them directly from within a chat, without needing other apps, websites, or extensions. This new feature, detected by WABetaInfo in version 2.26.35.3, is currently available to a very limited number of users in certain countries, but it points to a clear shift in Meta's strategy.

The idea is that these agents will appear as just another conversation within the chat list, with their own name and profile picture, allowing users to interact with them as if they were regular contacts. These aren't just typical chatbots: the platform supports both AI-powered assistants and simpler bots designed to perform specific tasks. The goal is for WhatsApp to act as a common interface for various external services, similar to what has already been seen in other parts of the world.

How to prevent being added to WhatsApp groups
Related article:
Personalized AI-powered responses on WhatsApp

How will agent connection work on WhatsApp?

WhatsApp AI agents

The setup process is done through the app's settings, where a new section called "Agents" is enabled . From there, the user can create an agent, assign it a name and a profile picture that will appear in the chat list. Each account can link up to five different agents, allowing for separate tools for specific functions without mixing them up. Once configured, WhatsApp generates an API key that must be entered into the service hosting the agent to establish the connection.

After completing the process, the agent operates within a one-on-one chat. The user can send messages and receive replies just like any other WhatsApp conversation. However, there's a significant limitation: the agent can only access messages the user shares within that chat. It cannot view other conversations, contacts, or files stored in other chats. Furthermore, these agents cannot currently be added to groups or communities, so all communication remains within one-on-one conversations.

The infrastructure for creating these agents can come from platforms and APIs of companies like OpenAI, Google, or Anthropic, although WhatsApp doesn't require the use of a specific provider. This means that, in theory, users could have Gemini or ChatGPT integrated directly into the application, although we've already seen cases where Microsoft Copilot is unavailable on WhatsApp . It's also possible to connect compatible, conventional bots, such as those already found on other messaging platforms. The difference between a bot and an agent lies in their capabilities: while a bot may be limited to responding to predefined commands, an agent can interpret instructions in natural language and perform more complex tasks on behalf of the user.

WhatsApp AI
Related article:
WhatsApp AI: the new tool to expand photos to 9:16 format

The fine print: no end-to-end encryption

WhatsApp AI agents

One of the most sensitive aspects of this new feature concerns privacy. Chats with third-party agents are not protected by end-to-end encryption , unlike regular conversations between people. The reason is technical: messages are routed through a service operated by Meta that connects the user to the agent provider, preventing the application of the encryption scheme used by the Signal protocol on which WhatsApp was built. To mitigate risks, it is advisable to review a comprehensive guide to protecting your privacy on WhatsApp.

In practice, this means that the agent's developer can access the content of messages sent to it as part of providing the service. WhatsApp also warns that AI agents' responses may contain errors and recommends verifying important information before acting on it. For users handling sensitive data, such as credentials, financial information, or trade secrets, this is a crucial point to consider before delegating any tasks to these assistants.

Agent management is also possible at any time through the settings. You can modify their name or image, disconnect them, and reconfigure them later. If an agent is deleted, the connection is immediately severed, and it becomes impossible to send or receive new messages, although the previous history remains visible in the conversation. To use it again, you would need to create it again and generate a new API key.

The context: regulatory pressure and Meta's strategy

This beta version doesn't arrive in a vacuum. WhatsApp's opening to external agents is partly a response to European regulatory pressure. In June 2026, the European Commission ordered Meta to restore free access to the WhatsApp Business API for rival AI assistants, after the company had blocked them in October 2025. Meta appealed, but the order remains in place until the conclusion of the broader antitrust investigation or until June 2029, whichever comes first. The beta's launch, with its own system of API keys generated by WhatsApp, appears to be the technical implementation of that mandate: instead of exposing the Business API, Meta channels the connection with external agents through a layer it controls.

On the other hand, this feature is part of Meta's broader strategy to transform WhatsApp into a conversational agent platform. In June 2026, the company globally launched Meta Business Agent, its own customer support agent integrated into WhatsApp Business and Instagram DMs. This bot already allows users to answer frequently asked questions, recommend products, schedule appointments, qualify leads, and escalate the conversation to a human if needed. Since August 1, 2026, Meta Business Agent has charged businesses $2 per million tokens, which is roughly equivalent to 4 to 5 cents per typical exchange.

The next layer is more ambitious. According to internal documents cited by The Information and reported by Forkast News and The American Bazaar, Meta is working on Project "Hatch," a consumer agent platform that would allow users to delegate complex tasks—such as placing orders on DoorDash, searching for products on Etsy, browsing Reddit, and managing their Outlook calendar—all from within the messaging interface. Hatch is reportedly inspired by OpenClaw, an open-source agentic tool created by Peter Steinberger in November 2025, which Meta internally banned in February 2026 for security reasons, shortly before Steinberger left for OpenAI. During the Q2 2026 earnings call, Mark Zuckerberg advocated for a capital expenditure commitment of between $125.000 billion and $145.000 billion for the full year, with a three-pronged strategy: maintaining advertising and recommendation systems, scaling business agents and APIs, and deploying consumer agents to its 3.500 billion users.

In short, the beta version that WhatsApp has just activated for a small group of Android testers is more of a preview than a full release. But it foreshadows a clear strategic move: Meta wants to make the world's most popular messaging app the default front-end for interacting with AI agents, for both businesses and consumers. For users, the relevant change isn't technical—the wiring is just an API key and a chat—but strategic: the place where they already chat daily will soon have an integrated AI layer, and the way we interact with digital services could change forever. The feature is being rolled out gradually, and Meta hasn't confirmed when it will be available to everyone, but it seems to be just a matter of time.


Add as preferred source in Google