OpenAssistantGPT

What OpenAssistantGPT is

OpenAssistantGPT is a chatbot and AI agent builder that sits on top of the OpenAI API. Its home page headline is "Build your own AI agent with no code", and the description is plain: a smart RAG agent for customer support, sales and similar jobs that you embed on your website so it can answer your users' questions. It is aimed at small support teams, founders who want a bot trained on their own site, agencies running bots for several clients, and developers who like that the project is open source. The company says it is "used by over 4000 people". The pages do not say where the company is based.

It also has a second identity that matters for this directory: it is open source. The home page says the platform "is open source and powered by open source software", and the code is on GitHub.

What you build and how

You build a retrieval-based chatbot. The pitch for the editor is short: "Give us the URL to crawl and we finish the job." You point the product at your website, a crawler reads it, and the bot answers from what it found. The home page shows this as a live "Crawling your website" demonstration, and the plans count crawlers and files as separate limits, so a bot's knowledge can come from the site it reads, from uploaded files, or from both.

The brain is OpenAI's Assistants technology. The page says "We use OpenAI Assistant to power our chatbots" and that you can use GPT-4, GPT-3.5 and GPT-4o. A banner on the home page announces two newer moves: an upgrade path "from Chatbots to AI Agents" and support for the OpenAI Responses API, described as faster and more reliable with improved tool calling. For a builder, that means the model choice and the tool-calling behavior follow OpenAI's platform rather than a model the vendor trains itself.

The feature list on the home page is the most concrete description of what a finished bot can do:

• User inquiry. A visitor can leave contact information with a question, so the bot collects leads.
• AI Agent Actions. The bot can query API endpoints with specific URL parameters, which lets it fetch dynamic data and respond with it.
• Web Search. It can query the internet for the newest information.
• File attachments. A visitor can attach a file such as CSV, XML or an image in the chat, and the bot analyzes it.
• SAML or SSO authentication. You can create a private chatbot that only your employees can use, which turns the same builder into an internal assistant.

The embedding story is broad. The home page promises integration with "the most popular website platforms", and a dedicated integrations page lists no-code and low-code website builders whose custom HTML option accepts the bot, including Webflow, Showit, Bubble, Jumpseller, WebWave and Bootstrap Studio, among many others. So the end user of your site meets a chat widget; the person building it works in the OpenAssistantGPT dashboard, whose sections are named in the footer as Dashboard, Chatbots, Files, Actions and Custom Domains.

For developers, the open-source side is the escape hatch. The home page shows two repositories. The main one, OpenAssistantGPT/OpenAssistantGPT, is described as "An Open-Source SaaS Platform for Crafting Chatbots with OpenAI's Assistant", written in TypeScript, with 431 stars and last updated in 2024. The other, chatbot-sdk, is "A SDK for building and deploying your own chatbot using NextJS in Vercel", which you can then place in an iframe on your site, with 62 stars. The pages do not publish a self-hosting guide or name a license, so check the repository itself before planning to run your own copy.

Other things on the site's menu round out the picture: Documentation, a Changelog, a blog, an Enterprise page and an Agency page. The home page also announces xseek, a separate new product for getting a product mentioned by leading LLMs, which is unrelated to building a chatbot.

Plans, limits and claims

Pricing is on the home page, and the free plan is a genuine free plan, with no credit card required to start.

• Free Plan, $0. 1 chatbot, 1 crawler, 3 files, 1 action and 10 messages per month. The company describes it as suited to personal projects and experiments.
• Basic Plan, $18 a month. 9 chatbots, 9 crawlers, 27 files and 9 actions, plus customizations, unlimited messages, client inquiry and lead collection, and a chat disclaimer.
• Pro Plan, $54 a month. 27 chatbots, 27 crawlers, 81 files and 27 actions, plus 5 custom domains, the option to remove the "Powered by OpenAssistantGPT" line, and client file attachment. The company calls its whitelabeling options great.
• Enterprise Plan, custom. The page shows the limits as "X" placeholders and says to contact the team. It adds SAML or SSO chatbot authentication and an SLA guarantee, and notes that agencies managing multiple clients also use this plan.

The company's performance claims sit on the same page: a 35% reduction in support tickets, a 60% improvement in resolution time "from the GPT chatbot", and $25,000 in annual cost savings described as an average for businesses that reduce support staff requirements. These are the company's figures, with no study or customer named behind them.

Two limits deserve attention. The free plan's 10 messages a month is enough to test, not to run a live site. And the pages do not explain how the cost of the OpenAI model calls behind a bot is handled on the paid plans, so ask before putting a bot in front of heavy traffic.

Who it suits and who it does not

It suits a founder or a small support team that wants a chatbot trained on its own website with a free tier to start, and wants that bot to capture leads, call an API or search the web. Agencies get a plan with 27 chatbots and 5 custom domains for the price of $54 a month, which makes per-client bots plausible. A company that wants a private employee assistant behind SAML or SSO has a direct path through the Enterprise plan. Developers who like to read the code before trusting a tool have the repository, and the SDK for a NextJS deployment.

It is less suited to teams that want a model other than OpenAI's, because the pages describe only OpenAI models. It does not publish a compliance certification, a hosting region or a data-retention statement, so a regulated buyer has questions to ask. And it is a chatbot builder, not a general app or workflow platform: you will not find a visual database or a page designer here, only the bot and the places you embed it.