One data model is the AI substrate.Everything else is plumbing.
Kugamon does not ship an AI feature. It ships the thing AI needs: revenue records that live in Salesforce as native objects, under Salesforce's security model, reachable by Agentforce, by Salesforce-native AI tools and by any LLM connected through the standard APIs. This page shows what that looks like in practice, including the walkthroughs of Claude creating quotes, orders and renewals by voice.
AI agents act on data they can reach. Kugamon's revenue records — Quote, Order, Invoice, Payment, Contract, Subscription, Asset — are native Salesforce records in your org, on the standard Product and Price Book. To Agentforce, to a Salesforce-native AI tool, or to Claude connected through the standard APIs, a Kugamon Quote is just more Salesforce data: same objects, same security model, no integration work.
One unified data model is the AI substrate
There are three ways an AI agent can get at revenue data, and they are not equal. It can read native objects. It can read what a sync layer copied in from an external CPQ or billing tool, as of the last run. Or it can be wired to the external tool directly — one integration per tool, per use case, maintained forever.
Kugamon takes the first path by construction, which gives every current and future agent the same three properties:
- Agentforce works on day one. Native objects mean an agent reads and acts with no custom integration. Quote-to-cash data is Salesforce data to the agent. (Salesforce's own Agentforce Revenue Management is a separate product on a separate data model; the guide explains the difference.)
- Salesforce-native AI tools inherit the model. The same objects, the same security, the same reports — a native tool sees Kugamon records without Kugamon doing anything.
- Direct-to-AI apps work through native APIs. Claude, ChatGPT and Gemini reach Kugamon objects through the standard Salesforce APIs. There is no separate integration build and no middleware to keep alive.
The longer argument, with the failure modes of sync architectures, is in Why AI agents need native revenue data.
What Agentforce sees
Agentforce operates on Salesforce objects under Salesforce's security model. Because Kugamon holds no copy of your data anywhere else, the agent's view is complete and current by definition: the open Quote a rep is editing, the Order that was released an hour ago, the Invoice with a Balance Due, the Subscription whose IsActive flag the scheduled job just recomputed. Object permissions and field-level security apply to the agent as they do to the user — see Native architecture for the object list and Security for the model.
What you build is the agent: its topics, the actions it may take, its guardrails. That is Agentforce work, not Kugamon work, and it is the same work you would do for any object in the org.
The public skills library
An agent that can see records still needs to know the lifecycle: that a Quote is disposable and an Order is not, that Release is the event that creates the downstream records, that an Expansion co-terms to the Contract End Date. Kugamon publishes that knowledge as a skills library on GitHub, kugamon-skills, so any LLM agent can drive Quotes, Orders, Subscriptions, Renewals and Billing with structured guidance rather than guesswork.
It is part of a small open toolkit at github.com/kugamon: salesforce-core-skills, a set of general Salesforce admin and developer skills that work with any Salesforce MCP server; salesforce-isv-skills for partner workflows; and a Salesforce MCP server for Claude Desktop. They install as Claude plugin marketplaces, or via npx for other agent tools. The core skills and the MCP server are MIT-licensed; kugamon-skills is public but proprietary, because it is tied to the commercial packages. The toolkit guide covers installation and what each skill does.
Claude + Cirra AI: the walkthroughs
The series records Claude connected to a Salesforce org through Cirra AI's MCP server, with the Kugamon skill installed, running the lifecycle by voice in a real org. Every video is on the videos page; the sequence is:
Setup
- Connect Claude AI to Salesforce — Cirra AI MCP Server Setup for Kugamon (Admin Tutorial) — the admin gateway every other video assumes.
- Connect Claude AI to Salesforce in 5 Minutes (End User Setup) — each user activates it for themselves.
Records by voice
- Create a Salesforce Account with Claude AI — by voice, in about a minute.
- Create a Salesforce Contact with Claude AI
- Create a Salesforce Opportunity with Claude AI
Quotes
- Create a Quote in Salesforce with Claude AI — lines, pricing and discounts, in about a minute, because the quote is native.
- Update a Quote in Salesforce with Claude AI — 75 licenses instead of 50, re-priced, approvals re-triggered.
- Clone a Quote in Salesforce with Claude AI in 30 Seconds
- Cancel a Quote in Salesforce with Claude AI — dead quotes out of the forecast.
Orders
Create an Order in Salesforce with Claude AI — from a Quote, or by direct entry.
Clone an Order in Salesforce with Claude AI in 30 Seconds — reorders and multi-site rollouts.
Release an Order in Salesforce with Claude AI — the downstream cascade: Contract, Asset, Subscription, Invoice.
Cancel an Order in Salesforce with Claude AI — Clean Cascade Reversal
Hit Your Claude Usage Limit? Add More for Power Users — a practical note for heavy users.
The point of recording them in a real org is that nothing is staged. Claude is reading and writing the same objects your reports read, through your user's permissions.
What this is not
Kugamon is not an AI product, and this page should not read as one. The package contains no model, no assistant and no AI service, and Kugamon has no access to customer data. What it contributes is architectural: the data is native, the security model is Salesforce's, and the lifecycle is documented for agents in public. Whether an agent is worth deploying, and for which tasks, depends on your data hygiene and your process — an agent creating orders against accounts with missing addresses will fail the same way a rep does.
Where to start
Watch the two setup videos, then repeat the Create a Quote walkthrough in a 30-day trial org, which comes with the suite installed and admin rights. If you are evaluating vendors, ask each one where the data lives, whether an agent can act on it and not just read it, and what the vendor publishes for agents to consume — then ask for a live demonstration against your own org's security model.
Yes, in the sense that matters: Kugamon quotes, orders, invoices, payments, contracts and subscriptions are Salesforce objects in your org, so an Agentforce agent reads and acts on them under the same permissions as any Account or Opportunity, with no connector. You still build and scope the agent — its topics, actions and guardrails — the way you would for any Salesforce data.
A public GitHub repository, kugamon-skills, that teaches an LLM agent Kugamon's lifecycle — quotes, orders, order release, subscriptions, renewals, billing — with structured guidance on the objects, record types and sequence of steps. It sits on top of an open salesforce-core-skills plugin of general Salesforce skills and a Salesforce MCP server, and installs as a Claude plugin marketplace or via npx for other agent tools.
An admin connecting Claude to a Salesforce org through Cirra AI's MCP server and installing the Kugamon skill, then a rep creating an Account, a Contact, an Opportunity and a Quote by voice, updating, cloning and canceling quotes, creating, updating, cloning and releasing orders, and canceling an order with the full downstream cascade reversed — each in a real org, most in under five minutes.
No. Kugamon has no access to customer data and runs no AI service. Agentforce runs inside Salesforce; a Claude or ChatGPT session connected through Cirra AI talks to your org through your user's authorization and Salesforce's APIs, so the data path and the permissions are the ones you already govern.
Yes. Contracts, Subscriptions and Renewal Opportunities exist only with the Subscription Management add-on, so an agent can only act on them in orgs running it. The skills library detects which Kugamon packages are installed and routes to the right record types.
With the two setup videos — the admin one connects Cirra AI to the org and installs the Kugamon skill, the end-user one takes about five minutes per person — and then the Create a Quote walkthrough in a trial org. The 30-day trial comes with the suite installed and admin rights, so the whole series can be repeated on your own data.
What Salesforce admins say on AppExchange
Quoted verbatim from the Kugamon listing on AppExchange. Reviewer names and dates as published there. Read all 134, including the critical ones →
★★★★★A new bar for RevOps
“As someone who has set up Salesforce CPQ before, I found Kugamon to be much more straightforward and easier to use. It took us just a few weeks to get going, rather than the few months or even years required by other systems.”
★★★★★Easy to understand, easy to implement
“This SFDC-native platform is lightweight but powerful, bringing the entire sales process onto one pane of glass. No more disconnected Opportunities and Quotes/Orders, no more fractured approval processes, just more control over everything. Implementation is a breeze.”
★★★★★Clear & Intuitive
“The Kugamon team is incredibly responsive and handle support questions very quickly. It’s clear they genuinely value customer feedback and are always innovating.”
Configure, price, quote, sign, bill, collect, and renew.
All natively in Salesforce. Deploys in weeks, not months, usually without a systems integrator.