Your ERP · 4 in stock · $2,840Luis Ortega · Ortega's Kitchen · Source: WhatsApp · Opt-in: yes1 range · $2,840 · Stage: quote sentCall Luis · 10:30 a.m.No copy and paste · loggedInventory, orders, balances and accounts receivable, looked up and updated from the conversation.
Each conversation creates or updates the contact, the deal and the task.
Tickets opened, updated and closed with the customer’s details already filled in.
The agent asks your billing system for a payment link for ACH or card, and confirms the payment once your system posts it. Konnet doesn’t process payments or issue invoices.
Balances, payoff amounts and application status for lenders and credit unions.
Your in-house software, connected through its API.
An OpenAPI file, a collection export or a PDF from your vendor.
Our team uses Konnet Core, our internal tool, to read it and publish your system as an MCP server: a list of named actions your agents understand.
A real call shows you what each action returns before anything goes live.
That’s how long a documented API takes to connect. If your system has no API, we tell you before you commit.
When a prospect asks about a product on WhatsApp, check stock and price in our ERP, create the contact and deal in our CRM, send the quote and book a call. Discounts over our limit need the sales manager.
Done: 6 steps. 1 waits for a person's approval and the rest run on their own.
| Step | What the agent does | What a person approves |
|---|---|---|
| 1. A prospect asks about a product | Understands the request and the quantity | |
| 2. Checks stock and price | Looks them up in your ERP by direct code | |
| 3. Creates the record | Creates or updates the contact and the deal in your CRM, with the opt-in | |
| 4. Sends the quote | Asks your system for the quote and shares it in the chat | |
| 5. Prospect asks for a discount over your limit | Prepares the request with the deal details | The sales manager approves or counters |
| 6. Books the call | Creates the task for the rep on the account |
This flow is built in Konnet Flow: you describe the process, the Copilot builds it and you approve it before it goes live.
See Konnet FlowFor companies in New York, Los Angeles, Chicago, Houston, Miami and across the United States: setup and go-live happen remotely.
WhatsApp requires an opt-in before you message a customer, and rules such as the TCPA may also apply. The agent can write each opt-in, with date and source, to the contact record.
A “stop” on WhatsApp updates the CRM so no other flow messages that person.
Every operation in your systems is logged with who, what, when and on which data.
Each agent works with the role you assign: look up only, or also log a payment or create an order. Sensitive actions wait for a person.
Your system’s credentials are encrypted, isolated per company and never appear in a log.
Checking a balance or creating an order runs as direct code against your system. AI only steps in to understand a message or read a document.
Every connected system is published as a standard MCP server for your agents in Konnet Flow.
Using those tools from agents your team builds, or through Konnet’s REST API and webhooks (payment received, document validated, ticket created), is enabled on request.
Each new source is set up by configuration on the same platform, not a development project.
AI agent integration connects an AI agent to the software your company already runs, like your ERP, CRM, help desk or billing system, so the agent can look up real data and take action instead of only chatting. It matters because the value of an agent is the work it finishes: an order logged, a ticket closed, a payment confirmed. With Konnet, every action has a permission and a log, and sensitive steps wait for a person on your team.
Any system with an API: ERP and accounting, CRM, help desk, ecommerce, billing, loan servicing and software built in-house. Konnet doesn’t depend on a closed list of brands; each new connection is configuration on the same platform. If your system publishes API documentation, our team reads it and sets up the connection, in about 5 minutes when the documentation is complete. If your system has no API, we tell you before you commit, so there are no surprises.
No. Many companies look for a WhatsApp CRM because their team answers from a phone and loses track of follow-ups. Konnet connects WhatsApp, Instagram, Messenger, web chat and email to the CRM you already have: each conversation creates or updates the contact, the deal and the task, and your team keeps working where it always has. The agent qualifies the prospect, answers with real data and books the next step on the right rep’s calendar.
If your system has a documented API, our team reads the documentation with Konnet Core, our internal tool, and has the connection ready in about 5 minutes, without writing code. Then you decide what the agent may look up and what it may update, and we configure it. In the demo we connect a real source so you can see the agent work with your own data. If the documentation is incomplete, we tell you exactly what to request from your vendor.
A custom system connects like any other as long as it exposes an API, even an internal one. Our team publishes it as an MCP server so the agent understands what it can look up and what it can update, with clear limits. Konnet needs an API to connect a system. If yours has none, we review the situation with your IT team before committing to a date, and tell you plainly what can be automated today and what can’t.
You decide. Each agent works with the role you assign: it can stay read only, checking balances, orders or inventory, or it can also update, like logging a payment or creating an order. What isn’t allowed, it doesn’t do. For sensitive actions, like a refund above a certain amount, you set a limit and the action waits for a person’s approval. Every operation is logged with who, what, when and on which data, so you can audit it like any employee’s work.
No. With Konnet, checking a balance, creating an order or logging a payment runs as direct code against your system, through its API or MCP server, without spending tokens. AI only steps in where it adds value, like understanding what a customer wrote or reading a document. Because each step gives the AI one narrow task, low-cost models are enough. That’s why pricing is per resolved action, USD 0.25, and not per token or per message.
Yes, on request. Konnet publishes each connected system as a standard MCP server for your agents in Konnet Flow. If your technical team wants to use the same tools from agents it builds, or through Konnet’s REST API and webhooks, we enable it on request, with the same permissions and audit trail. That way you keep a single set of connections and rules, whether the agent asking is one of ours or one of yours. See the MCP server page for details.