Pitch sheet

Sell AI assistants under your own name.

For agencies first, then freelancers, software partners and developers. One engine you run yourself, a private assistant for every client, English, Hindi and Hinglish. Print this page or share it on WhatsApp.

Try it free with one client Read the step-by-step guide

01The 30-second pitch

For a digital marketing or web agency. Say it in this order.

Who
Digital marketing and web agencies, and freelancers, whose clients keep asking for an AI chatbot.
Problem
Clients want an assistant that answers from their own documents, and you cannot risk one that makes things up. Website projects end, and the income ends with them.
What it does
cob-rag-engine is a finished platform you run on your own server: a separate private assistant for every client, a website widget with leads, answers on the client's own WhatsApp number, English, Hindi and Hinglish, plans, prepaid balances and Razorpay payments with GST invoices. Your name, logo and colours on the dashboard and e-mails.
Proof
412 of 412 isolation attacks refused in the repository's own test. Every price, date and phone number in an answer must appear in the client's documents, or the assistant says it does not know. Search measured on six public datasets (mean nDCG@10 0.537 against 0.526 published for the same embedding model, bge-base-en-v1.5, measured before the default became bge-m3). All numbers are in the benchmark report.
Ask
Try it free with one client. Write to callobuzz@gmail.com, and install it yourself or ask us to help set it up.

How you earn. Charge each client a monthly fee for their assistant. You set the price, in ₹ or $. Plans, caps and usage counted per client keep the bill fair. Your costs are the server and the AI calls (a client can bring their own AI key). What you earn depends on your price and your clients; no income is promised here.

02How it works

You run one engine. Each client gets their own private assistant on it.

  1. Install it, or ask for help. Docker on your own server; the guide is step by step.
  2. Add a client's documents. PDF, Word, Excel, web pages and scans, including Hindi scans.
  3. Put it on their website. One script tag, locked to their domains, with a button for hand-off to a person on WhatsApp, phone or e-mail.
  4. Bill them every month. Usage is counted per client, so the invoice is easy to explain.
How cob-rag-engine fits between a client's documents and their website A client's documents go into your engine on your own server, which keeps a private knowledge base per client and checks every answer. The answer appears in a chat widget on the client's website, with a WhatsApp hand-off. The client pays you a monthly fee. CLIENT'S DOCUMENTS PDF · Word · Excel · web pages scans, including Hindi upload YOUR ENGINE · YOUR SERVER A private knowledge base per client Search + fact check on every answer Plans, caps and usage per client English · Hindi · Hinglish answer + sources CLIENT'S WEBSITE Chat widget · one script tag leads · WhatsApp hand-off MONTHLY fee to you You set the price. Your name on the dashboard.
Blue is your server, amber is the client paying you, cyan is data moving.

03Why not LangChain, LlamaIndex, Dify or an agent SDK?

Say it in four lines.

  1. LangChain, LlamaIndex and the agent SDKs give you parts to build with; cob-rag-engine is the finished knowledge platform: customers, keys, documents, fact-checked answers, a website widget and billing, already built and tested.
  2. They are not rivals: your LangChain, LlamaIndex, OpenAI or Claude agent plugs into it over MCP and uses it as its governed knowledge base.
  3. Dify is a full platform too, but it is built for designing apps, and its licence needs written authorization to run it for many customers; cob-rag-engine is built to resell, with plans and billing, Hindi and Hinglish, WhatsApp and DPDP consent, and answers that stay inside the customer's own documents.
  4. So you sell assistants from day one instead of building the platform under them, and if all you need is a one-off prototype, LangChain alone is the right call.

The longer comparison, with a table and diagrams, is on the landing page.

04Not an agency?

The same engine, three other ways to use it.

Freelancers and first-time founders

Start your own AI chatbot business this week. Your brand, your prices.

  • A finished product: customers, documents, fact-checked answers, widget, dashboard.
  • Open source under AGPL-3.0; you can charge your clients, and a commercial licence is available on request.
  • Prepaid wallets in ₹ or $; price per question, or AI cost plus your markup.
Software partners

Add an AI assistant to what you already sell to the same clients.

  • A separate engine with its own API; nothing changes inside your product.
  • Sync a client's catalogue or help centre through the records API.
  • Usage counted per client, so the add-on is easy to price.
Developers building agents

Plug trustworthy company knowledge into your agents.

  • cob-rag-mcp: 52 tools for any MCP client, plus REST and Python and JavaScript clients.
  • A read-only named key per agent.
  • Cited answers, or an honest "I don't know".

Try it free with one client

cob-rag-engine is open source under AGPL-3.0: get the code. There is no hosted cob-rag-engine: you run it on your server, and we can help you set it up.

Callobuzz · callobuzz@gmail.com callobuzz.github.io/cob-rag-engine
E-mail us

Example businesses shown on this site are made up. Numbers come from the benchmarks in the repository and can be re-run. WhatsApp answers need the client's own WhatsApp Business number and a Meta app. Income depends on your own prices and clients and is not promised.