Last updated: June 2026
TL;DR
On Invent, every model comes with built-in tools: web search, file scraping, reasoning, image generation, and research. An Auto layer picks the right tool and model for each message, so you get agent-level capability through a simple chat, while staying in control of what it can do.
Most "AI chatbots" can only do one thing: talk. They answer, and the moment a task needs an actual action, looking something up, reading a file, making something, they hand the work back to you.
Invent takes a different approach. You still talk to it the simplest way there is, plain language in a chat. But under the hood, it works like an agent: every assistant comes with built-in tools and does the heavy lifting for you. The most accessible interface on the surface, a capable agent underneath.
Better still, you do not have to manage any of it. The platform decides which tool fits the moment, you stay in control of what the assistant can use, and you get real capability with less to think about. Here is how it works, under the hood.
What are built-in tools?
A plain language model can only produce text. Useful, but limited. Built-in tools are extra abilities attached to that model so it can act in the real world: look something up, read a file, make a picture.
On Invent, every assistant gets the same core toolset today, whichever AI model is behind it:
- Web Search: pull current information from the internet.
- Scrape & Download: grab content or files from a URL: web pages, PDFs, images.
- Think: slow down and reason step by step before answering.
- Generate Image: create an image from a text description.
- Research: run a deeper, multi-step investigation across sources.
Because the tools live on the platform and not inside one model, you can switch models anytime and keep every one of these abilities.
The tools your assistant can use today
Here is what each tool unlocks, and a real reason a business owner would want it.
Web Search
Your assistant can look up live information instead of relying only on what its model was trained on. Useful for questions about current prices, availability, or anything that changes. Prefer answers grounded strictly in your own knowledge base? Turn it off. The choice is yours.
Scrape & Download
Point the assistant at a URL and it can read the page or pull the file behind it, including PDFs and images. Handy for turning a supplier's spec sheet, a help doc, or a product page into knowledge your assistant can use right away.
Think
For anything that needs care, the assistant reasons through the problem step by step before it responds. Fewer rushed, surface-level answers on the questions that actually matter.
Generate Image
Describe what you want and the assistant creates it: a product mockup, a simple social graphic, an illustration. No separate design tool, no leaving the chat.
Research
For bigger questions, the assistant runs a deeper, multi-step investigation: it gathers and cross-checks across sources, then brings back a structured answer instead of a single quick reply. Useful when you need a real briefing, not just a one-line response.
The "Auto" layer: less for you to think about
The quiet hero is the Auto layer. It reads every message and decides, automatically:
- Whether a tool is needed at all, and which one to call.
- Which model is best for the job, switching if the primary one cannot handle the request. If a tricky file type trips up one model, Auto falls back to a more capable one that can read it.
This is the part that keeps capable AI from becoming a chore. You do not have to learn which tool does what, or which model is best for which task. Auto handles the busywork and picks the combination that just works. (Prefer to drive it yourself? You can still pick the model per assistant, from the latest GPT, Claude, Gemini, and Grok.)
The Auto layer picks the right model for each task, across GPT, Claude, Gemini, and Grok.
You stay in control
Capability is only comfortable when you stay in charge of it. So the tools are built around a simple rule: the assistant can be as capable, or as restricted, as the task calls for, and that line is always yours to draw.
You decide whether web search is on or off. You decide whether answers come from the open internet or only your own knowledge base. As the toolset grows, that same principle holds: for anything that can reach outside your workspace, you set how much access it gets. Powerful AI, with a human in control. It is the idea behind everything we build.
Where we're headed
Built-in tools are the start, not the finish. The direction is an assistant that does not just answer, it operates, while staying every bit as simple and controlled as it is today.
What we are building next (in development):
- Computer: a sandboxed workspace where your assistant builds real artifacts: charts, documents, spreadsheets. The goal is simple, ask for a chart of your top products by sales, and instead of explaining how to make it, the assistant builds the chart and hands you the finished file.
A preview of the Computer tool (in development): it can even generate an image with no internet access, network set to Off.
Another Computer preview (in development): ask for a chart and it writes the code, runs it, and hands back the finished chart.
This is not live yet. We are building it the same way as everything above: powerful under the hood, light on your attention, and always under your control. We will share it as it is ready.
Why this matters for your business
A chatbot that only chats hands the work back to you. An assistant with tools does more of the work and hands you the result, without making you manage the machinery behind it.
Here is the difference in one small task. Ask a generic chatbot for a list of users as a spreadsheet and it prints the rows as text, and you still have to copy them, paste them somewhere, and save the file yourself. Ask Invent and it hands you the finished CSV, ready to download. Same data, but Invent does the last mile.
Ask for a CSV and Invent delivers the finished file, ready to download, not just text to copy.
That is the real shift: from a place your customers ask questions to a place where work gets done, across every channel you run, with you deciding how much the AI is allowed to do. Capable enough to be useful, simple enough to trust.
FAQs
Can Invent's AI search the web?
Yes. Web Search is a built-in tool on every assistant, so it can pull current information from the internet. You can also turn it off when you want answers grounded only in your own data.
Can Invent's AI generate images?
Yes. The Generate Image tool creates images from a text description, right inside the chat, no separate design tool needed.
Can Invent's AI do in-depth research?
Yes. The Research tool runs a deeper, multi-step investigation across sources and returns a structured answer, rather than a single quick reply, when you need a real briefing.
Do the tools work with any AI model?
Yes. The tools live on the platform, not inside one model, so every assistant gets the same toolset whichever model powers it, GPT, Claude, Gemini, or Grok, and the Auto layer can switch models without losing any abilities.
Do I have to choose which tool the AI uses?
No. The Auto layer decides which tool and model fit each message, so you do not have to. If you would rather control it, you can set the model per assistant and toggle tools like web search on or off.
Can I control what the assistant is allowed to access?
Yes. You decide whether web search is on, and whether answers draw from the open internet or only your own knowledge base. As more capable tools arrive, you will set how much access each one gets.
Can the AI build charts or documents?
Not yet. A Computer tool that builds charts, documents, and spreadsheets is in active development. We will share it with users once it is ready.
