ChatFrom
In development
Agents built from your own knowledge

An agent that knows your product.

ChatFrom is an AI chatbot for small teams, built from your own documents and website and put on your site with one script tag.

It answers from your documents and website with the source, asks your customers for what it needs, books calls and creates records through your own backend, and hands over to a person when it should.

In developmentSee how it works

Sign-up opens soon

STARTvisitor messageKSKnowledge SearchsearchCMChat ModelanswerCConditionintentanswercallAtAsk the VisitorphoneTCTool Calllist_callback_slotsAtAsk the VisitorwhenTCTool Callrequest_callbackSMSend Messageconfirm

Each step lights as the agent runs it.

SSamantha · Acme AnalyticsOnline
Ask a question…
01Answers cite their sources
02OpenAI, Anthropic, Google, DeepSeek and more
03Each agent's knowledge kept apart
04Open-source engine
Use cases

Any kind of chatbot. An assistant for your customers, available 24/7.

Whatever your business does, the agent learns it from your own documents and website. A support desk, a sales assistant, an in-app helper, a booking clerk, an internal HR bot: the same agent, shaped by what you give it. It answers questions, does the tasks and hands over to your team, at any hour.

01

Customer support

Answers from your help centre and policies, cites where each answer came from, and passes the hard ones to your team.

What's your refund window?30 days on unused items.
02

Pre-sales and pricing

Explains plans and features from your pricing page and docs, and books a call when a prospect wants to talk.

Is Pro worth it for two people?Here's the difference, and I can book 15 minutes with sales.
03

In-app assistant

Embedded in your signed-in app, it knows who the user is and can list their projects or create one through your own API.

Create a project called Demo.Demo is ready in your dashboard.
04

Bookings and requests

Collects a phone number, an address or a choice in the chat, and books a time from your real availability.

Can someone call me back?Wed 16 Sep at 10:00 works. Booked.
How it works

Build, shape, deploy, improve.

KnowledgeAdd source
ALLFILESWEBSITEQ&A
pricing.mdindexed
help-centre (42 pages)indexed
returns-policy.pdfindexing

Drop PDFs, Markdown or text

or crawl a website or sitemap

KNOWLEDGE USED

8.4 MB of 50 MB

Where it runs

One chatbot, everywhere your customers are.

Build it once. It answers with the same knowledge, tools and workflows in each place, and every conversation lands in the same log.

  • Your websiteA chat bubble on any page, added with one script tag.
  • Inside your productIn your signed-in app, knowing the user and acting for them.
  • Share linkA full-page chat to send by email or post anywhere.
  • TelegramA bot people message, in a private chat or a group.
  • SlackAn app in your workspace, in messages and threads.
  • DiscordA bot in your community's server, and in direct messages.
  • CrispAnswers in the chatbox you already have, and hands over to your team.

Setting up a chat app? Read the guide →

Automation

Automation you draw, step by step on a canvas

Drag steps onto the canvas and connect them: search the knowledge, ask the visitor, call your API, hand over to a person. Conditions branch on the answer.

ADD STEPKnowledge SearchChat ModelConditionTool CallSend MessageHuman HandoffAsk the VisitorEnd
STARTVisitor message
CConditionintent
What does the visitor want?
createcallFALLBACKother
AtAsk the Visitorproject_name
Ask its name, keep as vars.name
TCTool Callcreate
Call create_project
CMChat Modelcreated_reply
Tell the visitor it is ready
AtAsk the Visitorphone
Ask for a phone number
TCTool Callslots
Call list_callback_slots
AtAsk the Visitorwhen
Offer the open times as buttons
KSKnowledge Searchsearch
Find matching passages
CMChat Modelanswer
Answer, citing the sources
EEndend
Finish the turn
Drag from a step's right handle to the next step's left handle. A condition has one handle per answer.
Channels

Build once, put it where your customers are.

acme-analytics.com
PRICING
CFSamantha
Pro is $29 a month with 10× the tokens. Want the full comparison?Yes please
Security and reliability

What your visitors never see, and what your team always does.

Built in from the first agent, on the Free plan too. Nothing to switch on.

01

Your data stays yours

Each agent's knowledge is kept apart and answers only that agent; delete it any time.

02

Signed-in users stay private

The user's token goes only to your own server, never to other tools or logs.

03

Allowed domains

The widget runs only on sites you list.

04

Answers checked

Phone numbers, choices and links are validated before your tools see them.

05

Limits built in

Rate limits per visitor and per address; monthly token caps.

06

No errors shown to visitors

Failures become a polite message; the reason goes to you.

Pricing

Start free, stay free while you try it.

Free

IN DEVELOPMENT

$0

  • Unlimited agents
  • 500K model tokens a month
  • 50 MB of knowledge
  • All features
  • No card
In development

Pro

COMING SOON

$29 / month

  • 10× the tokens, 10× the knowledge
  • 5M tokens a month
  • 500 MB of knowledge
  • Longer messages and answers
  • Higher rate limits
Coming soon

Evaluations run on your own OpenRouter key.

Questions

The questions people ask first.

01

What is ChatFrom?

ChatFrom is a web application for building an AI chatbot from your own documents and website. You add your knowledge, set how the agent talks and which of your tools it may use, and put it on your site with one script tag or share it as a link. It is made by DeadPixel Studio.

02

What does it cost?

The Free plan gives 500,000 model tokens a month and 50 MB of knowledge, with no card needed. Pro, at $29 a month, is coming and is not on sale yet. Registration opens soon.

03

Which AI models does it use?

Any model on OpenRouter: OpenAI, Anthropic, Google, DeepSeek, Mistral and more, with live prices and the cost of every answer shown. Evaluations run on your own OpenRouter key.

04

Where do my documents go, and who can see them?

Your documents are indexed for your agent only. Each agent's knowledge is kept apart and answers only that agent, and you can delete a source, or the whole agent with everything it knows, at any time. The engine that produces the answers is open source.

05

How do I put the chatbot on my website?

Paste one script tag from the Deploy page; the widget runs only on the domains you allow. There is also a share link that opens the chatbot as a full page, for an email or a menu.

06

What happens when it cannot answer?

It says so, in the words you set, and can hand the conversation to a person by email, Slack or a ticket in your own system. Visitors can rate every answer and say what was wrong, and you turn a weak answer into a question-and-answer entry the agent uses next time.

An agent that knows your product, and can act on it.

In developmentSee how it works