Frequently asked questions

Chatfarm is a platform that helps organizations build, customize, deploy, and manage their own AI assistants using their knowledge.

Instead of building AI infrastructure from scratch, your organization can create an assistant, configure its behavior, connect supported knowledge sources, and publish it for the people who need it.

Use it to help students find academic information, help employees navigate internal procedures, answer customer questions, or make organizational knowledge easier to access.

Chatfarm gives an organization a controlled AI assistant that can use its authorized knowledge, follow its access rules, and be deployed to its own audience through a website, subdomain, or custom domain.

Your organization already has valuable information. It may be spread across documents, websites, policies, help centers, and internal knowledge bases.

The problem is that people often struggle to find the right information when they need it.

Chatfarm helps turn that information into an assistant that people can question in natural language.

Instead of searching through dozens of documents or waiting for someone to respond, users can ask a question and receive an answer based on information made available to the assistant.

General-purpose AI assistants serve many users and purposes. Chatfarm focuses on helping organizations deploy and manage assistants tailored to their own needs.

Your organization can configure its assistant around its knowledge, instructions, access rules, and intended use.

You can also publish the assistant through a Chatfarm subdomain or a supported custom domain.

The distinction is not that general-purpose AI tools cannot work with organizational information. It is that Chatfarm is purpose-built for managing and deploying organization-specific assistants to defined audiences.

Chatfarm can support organizational use cases such as:

• Universities helping students find academic information.
• Schools answering questions about programs and policies.
• Companies making internal procedures easier to find.
• Customer-support teams answering questions from approved documentation.
• NGOs helping people navigate programs and services.
• Public institutions making official information more accessible.
• Businesses helping customers understand their products and services.
• Professional organizations providing information from approved resources.

The right use case depends on your organization's knowledge, audience, and requirements.

Your assistant can use the knowledge sources supported by your Chatfarm workspace.

These include uploaded files (PDF, Word, PowerPoint, spreadsheets, text, HTML, JSON, and images), public web pages, website or sitemap crawls, a captioned YouTube video, and connected accounts — Drive, Notion, Slack, GitHub, Zendesk help center, HubSpot, Linear, and dozens of other tools. You pick the files or pages to import. You can also authorize live websites for search without importing every page.

The goal is to make relevant organizational knowledge available through one assistant rather than forcing users to search through multiple disconnected sources.

Not necessarily.

Chatfarm supports document uploads and public website crawling, including sitemap import. From Integrations you can sign in to Drive, Notion, Slack, GitHub, Zendesk, HubSpot, and many other accounts, then choose which folders, pages, or articles to import. A running import can be paused or stopped.

Imported web pages refresh on a schedule you choose. Connected accounts can re-sync about once a day when auto-sync is on. Uploaded files update when you replace them. Synchronization is periodic, not instant, and connecting an account does not import every file by itself.

Yes. Integrations lets you sign in to Google Drive, Notion, Dropbox, OneDrive, SharePoint, Confluence, Slack, GitHub, Zendesk help center, HubSpot, and many other accounts. After you sign in, pick the folders, files, or pages to import — connecting does not index the whole account. You can pause or stop a running sync. Selected Google Docs export as text. Selected PDFs and Office files from Drive or OneDrive import the same way as an upload, within the usual size limit.

Published help-center sites — including Zendesk, Intercom, Help Scout, GitBook, Freshdesk, Document360, and published Notion hosts — can also be crawled from the address visitors already open, without signing in. That imports the articles people already read. It does not open the ticket inbox, create tickets, or refund orders.

No.

Connecting an account does not mean every document or record should automatically become available to every assistant.

Organizations choose which sources are added and which assistants can use them. This is especially important when a workspace contains confidential documents, internal procedures, customer records, or information intended for a limited audience.

An assistant should only use information made available through the applicable configuration and permissions.

Chatfarm retrieves relevant information from the knowledge available to an assistant and uses it to help generate a response.

This helps the assistant answer questions using your organization's materials rather than relying exclusively on a model's general knowledge.

Answer quality depends on the source material, document processing, retrieval, configuration, and model behavior.

AI can still make mistakes, so important information should be verified when necessary.

Yes, where the reply uses retrieved passages. Retrieved passages are numbered, and the assistant is instructed to cite them in the answer. The chat shows the sources that appear in that reply.

This makes it easier to inspect the document or page behind an answer rather than relying on an unexplained response.

The availability and precision of source references depend on the source type. Citations make replies inspectable; they do not guarantee every interpretation is perfect.

Yes. You can set the assistant's name, avatar, instructions, model, temperature, and preferred language (auto, English, or French).

Experience modes include Ask, Learn, Quiz, Practice, Research, Support, and Execute. You also control the welcome message, knowledge policy, and how the assistant behaves when sources do not support an answer.

For example, a university assistant may use a formal academic tone, while a customer-support assistant may use a conversational style. The purpose you write is the rule.

An assistant should not invent an answer simply because someone asks a question.

You configure fallback behavior: the assistant can stay within its knowledge, admit that it does not know, or refuse to guess. Support mode can offer a handoff to a person.

This is especially important for academic, institutional, and customer-support use cases, where an invented answer is worse than a clear limit.

Yes. An assistant may be public, private to organization members, password-protected, invite-only, or limited to email domains you allow. You can also require a Chatfarm sign-in before chat, invite specific emails, and block addresses.

These controls decide who can open the assistant.

Access to the assistant and permission to use individual knowledge sources are separate. Both must be configured when private information is involved.

Yes. People can open a Chatfarm path, a Chatfarm subdomain, a connected custom domain, or a floating widget on your site.

This gives your organization a recognizable destination for its assistant instead of requiring everyone to use a general-purpose AI application. Widget hosts can be limited to allowed website domains.

Yes. A Chatfarm organization can run more than one assistant.

For example, a university could have separate assistants for admissions, academic programs, student services, and internal administration.

A company could have assistants for customer support, employee onboarding, product documentation, and internal policies.

Separate assistants can be configured for different purposes and audiences, helping organizations avoid forcing unrelated use cases into one generic chatbot.

Yes. Chatfarm treats assistants as organization resources, not isolated personal projects.

Organization roles include owner, admin, manager, editor, member, and viewer. Team invitations and role changes live on the team page. Authorized people can manage assistants, knowledge sources, and configuration according to those roles.

Yes. Customer support is an important use of Chatfarm.

An assistant can help customers find answers from approved product documentation, policies, and other knowledge you attach. Support mode tells the assistant to resolve from that knowledge and offer a human handoff without inventing ticket numbers.

Concierge handoffs notify managers in Chatfarm. Creating tickets in Zendesk, Freshdesk, Gorgias, Help Scout, Intercom, or Salesforce is not available. Knowledge-based answers and live help-desk actions are different capabilities.

Chatfarm works with knowledge you upload or import, with the connectors listed above, and with optional tools. Outbound MCP lets ChatGPT, Claude, or Cursor search this assistant's knowledge. Inbound MCP tools in Execute mode can call an HTTPS server you connect.

Reading a document, looking up a current transaction, and modifying a business record are not interchangeable. Chatfarm does not currently run payment, calendar, or shop actions in Stripe, Shopify, or Calendly, and it does not create help-desk tickets.

Only supported operations should be enabled for each connection.

No.

Chatfarm lets organizations create and operate assistants without building an AI model or maintaining the entire AI application infrastructure themselves.

You configure the assistant, provide or connect its knowledge, and deploy it through Chatfarm. The platform handles the application and knowledge-management workflow, while supported AI models generate responses.

No.

Chatfarm is built so authorized people can create an assistant, add knowledge, set access, and publish it without implementing a retrieval system themselves.

Some connectors require administrator authorization or initial technical configuration, depending on the connected service.

Chatfarm separates customer workspaces. Organizations control who can manage assistants, which sources those assistants can use, and who can open a published assistant.

External connectors require careful permission management. Connecting a service should not automatically make its private information accessible to every assistant or visitor.

Review the permissions and data-handling requirements relevant to your use case before connecting sensitive information. These controls are not a security certification or a claim of regulatory compliance.

No AI platform can responsibly guarantee that every generated answer will be correct.

Chatfarm helps assistants use relevant organizational knowledge, follow configured instructions, and respond appropriately when information is insufficient.

Results depend on the quality of the source material, retrieval, configuration, and model behavior.

For critical decisions, organizations should establish appropriate review procedures and verify important information against authoritative sources.

The goal is to make the process straightforward.

An organization can start by creating an assistant, adding supported knowledge sources, configuring its behavior, testing responses, and publishing it for its intended audience.

The time required depends on the amount of content, the complexity of the configuration, and any required connectors or access controls.

A simple assistant may be much quicker to launch than a complex institutional deployment involving numerous sources and permissions.

Chatfarm is designed for organizations that want to make their knowledge more useful through AI.

This includes schools, universities, companies, NGOs, public institutions, ministries, and other organizations whose employees, students, customers, or communities need access to information.

The platform is intended to support different organizational sizes and use cases through configurable assistants, knowledge sources, and deployment options.

Start by identifying the information your audience needs and the questions your assistant should answer.

Then create an assistant, add supported knowledge sources, configure its instructions and access rules, and test it with realistic questions.

Once the answers and permissions meet your requirements, publish the assistant through a Chatfarm address, a connected custom domain, or a website widget.

For larger or more complex deployments, discuss your organization's requirements before connecting sensitive systems or launching to a broad audience.