You can still connect your own AI agent to Kommo chats — but not directly. Replies reach the chat through amojo, and Kommo has closed off the access: the channel for sending messages is now granted only on request to support. That's why "wire up n8n in an evening" guides no longer work for a new integrator. Three working paths remain, and they differ by price, control, and whose AI ends up replying to the customer.
What closed: amojo and channel registration
First, the fact that breaks half the guides. Kommo's official n8n integration works only with CRM data — it creates and updates leads, fields, deal stages. It does not send or receive chat messages, and Kommo's documentation says so plainly.
An AI reply is a chat message. And it doesn't go out through the regular API but through amojo — Kommo's chat transport. To send anything through amojo you need a registered channel: an identifier and a secret key. And here's what changed: you can no longer spin up the channel yourself. It's granted on request to Kommo technical support, with review — the registration procedure is in the developer documentation.
The date when self-serve registration was closed Kommo did not announce publicly. We'll say it as it is: when we registered our channel in early 2026, the "do it yourself" path already didn't work — only a request and a wait. For a solo integrator, that's the wall the "wire up n8n in an evening" idea runs into.
Below, the three ways to get around that wall — legally.
Path 1: Kommo's native AI agent
The shortest path is to connect nothing of your own and use Kommo's built-in AI agent. It's assembled in the interface, with no servers or channel registration: the channel already belongs to Kommo. For a small team that needs a typical FAQ responder, it's often enough.
Two caveats that change the math. First: since June 1, 2026 the native agent is out of the Base plan — it's now paid in separate packages. Second and more important: it runs on Kommo's model. You configure the behaviour, but you don't put in your own model, your full prompts, or your logic in n8n. If the task is a typical FAQ, that's no problem. If you need an agent with your script, your knowledge base, and external tools, the ceiling shows up fast.
Who it's for: small businesses that need a quick turnkey responder and don't mind a model that isn't theirs.
Path 2: your own agent via a registered bridge
The second path removes exactly the thing everyone trips over: channel registration. A ready-made bridge already holds a registered amojo channel, and you connect your agent to it instead of passing support review for every project. The flow is simple: the customer's message arrives at the bridge, the bridge hands it to your agent in n8n or straight to the model, the agent returns the reply, and the bridge sends it to the chat through its channel.
That's how our service AnyLinga works — let's be honest, it's our product. It sits in Kommo's message flow and passes the conversation to the external AI you configured. The AI stays yours: your prompts, your model, your logic in n8n. The bridge doesn't invent replies — it transports them. The bonus of the same pipe is translation: the customer's message and the agent's reply can be translated automatically when the operator and the customer speak different languages.
We don't repeat the step-by-step setup here — a separate walkthrough of how to connect an n8n agent to Kommo via the bridge lives with the product. For choosing the path, one thing matters: option 2 gives you your own AI in Kommo chats without registering the channel and without your own transport server. You need a Kommo Advanced plan or higher — that's where custom widgets are available.
Who it's for: agencies and teams with custom logic, their own model, or multilingual sales.
Path 3: your own channel
The third path is to do it all yourself: register your own channel and send to amojo directly. On paper it's full control. In practice, the entry is precisely that closed door: without a request to support the channel won't come up, and workaround schemes are fragile and easily fall outside Kommo's rules. For most teams it's a dead end today; it makes sense only where there's an in-house dev team and a willingness to maintain the infrastructure.
Who it's for: almost no one, except large teams with their own development.
Comparison: which path to choose
| Path | Whose AI | Speed | Control & flexibility | Who it's for |
|---|---|---|---|---|
| Native Kommo AI | Kommo | instant | low — someone else's model, paid packages | SMBs, typical FAQ |
| Own agent via bridge | yours | days | high — your prompts, model, n8n, translation | agencies, custom, multilingual |
| Own channel | yours | slow | maximum, but the entry is closed without support | large teams with in-house dev |
Stripping out the detail: path 3 is closed for most, and the real choice is between 1 and 2. The native agent, when you need something quick and turnkey and don't mind paying for a model that isn't yours. The bridge, when you need your own AI, your own logic, and translation.
What this means for integrators
Closing the direct access isn't bad news for the integrator — quite the opposite. Before, a client could say "a friend built me a bot in n8n in an evening" and undercut the value of your work. Now "connect AI to Kommo" is a service again: you have to pick the path, register or take a ready channel, configure the agent, and get it to production state.
The bridge gives the integrator a comfortable position. The service takes on the transport and the channel, while you build the valuable part — the agent's logic in n8n, the knowledge base, the human-handoff scenarios — and resell that to the client, keeping your model and your margin. For the LATAM market, translation is a bonus: one agent serves customers in Spanish and Portuguese, and the operator replies in their own language.
For the business owner the takeaway is simpler: ask your integrator which path they propose and whose AI will be replying to customers — rented from Kommo or yours. That determines whether you'll be able to change the model and the logic later.
What about WhatsApp rules?
One restriction hangs over all three paths. Since January 15, 2026 Meta limits free AI chatbots on the WhatsApp Business API — and that doesn't depend on whether your agent is the native one, yours via a bridge, or your own channel. The bot's registration, its limits, and the scenario for handing the conversation to a human must meet the new rules. How to set that up step by step we cover in a separate guide on building a Meta-compliant AI bot in Kommo for 2026.
The bottom line: the direct "n8n → chat" path is closed — replies go through amojo, and the channel now goes through support. After that, three options: native Kommo AI for a quick turnkey, your own agent via a ready-made bridge for your logic and translation, and your own channel — closed for most. For custom and multilingual, the bridge is the pick.