Get started

Your first working quote, step by step.

You do not need to build an AI agent here. QuoteKernel gives your existing agent the business logic needed to quote. Start by teaching the kernel what your business sells.

  1. 01Request access

    The business you're configuring

  2. 02Choose what you quote

    Your industry model and services

  3. 03Enter pricing rules

    Train the kernel on your pricing

  4. 04Run test jobs

    Check it against prices you know

  5. 05Connect and verify

    MCP server or API

00 / Your path

Which one are you?

I run a business

Steps 2 to 5 can all be yours: what you sell, how you price it, checking the result, and connecting the agent you already use. Or hand step 5 to your agency or developer.

Teaching the kernel

I build agents for clients

Configure a kernel for each client, then connect every agent the same way.

For agencies

I am a developer

Step 5 and the docs: the tools, the states, and the rules for your agent.

Read the docs

01 / Step 1

1. Request access

Send an early access request. Tell us your name, work email, company, and what you build or quote for. We review each request and email you with the next steps.

02 / Step 2

2. Choose what you quote

Pick a ready-made industry model to start from. It comes with the job types, questions and common constraints for that kind of work.

Then list the services the business offers and the job types a customer might ask about.

03 / Step 3

3. Enter pricing rules

Train the kernel on how the business prices. Define:

  • the required inputs: the questions it asks a customer before quoting,
  • rates, packages and add-ons,
  • minimums and discounts,
  • exceptions: what it won't price,
  • approval rules: when a quote must wait for a person.

You can also load past quotes and invoices and review what QuoteKernel picked up before anything is used. QuoteKernel checks everything and tells you exactly what to fix.

What teaching the kernel involves

04 / Step 4

4. Run test jobs

Try a standard job, a job with a missing input, and an edge case. Use jobs you already know the price of. Correct anything that does not match your policy.

The guided scenario review walks you through it: mark results as right, too high or too low, and explain the exceptions.

05 / Step 5MCP · API · SDK

5. Connect and verify

Point your agent at the QuoteKernel MCP server, or use the API. Any MCP-capable agent works, including Claude. You can connect the chatbot, voice agent or workflow you already use yourself, or have your agency or developer do it.

Once connected, your agent sees the QuoteKernel tools: list the services, start an estimate, send the customer's answers, and get the estimate.

Run a complete quote from customer question to final state. Watch your agent:

  1. ask QuoteKernel what to ask,
  2. put those questions to you,
  3. pass your answers back,
  4. give you the price, say it's an estimate, and give you a reference number.

Connection details in the docs

Tell your agent the three rules

Add these to your agent's instructions:

Agent instructions
- Pass along what the customer said. Don't fill in answers for them.
- Never state a price QuoteKernel didn't return.
- Relay QuoteKernel's sentence word for word.

The tools tell the agent this too. Saying it twice helps.

Ready to make your agent quote?