Common questions about Agentivity

Short answers to the questions people ask first. Each answer links to the page that explains it.

A platform to build and run AI agents, AI teams and workflows, without writing code. You build in the Studio, a visual editor in your browser. See What Agentivity is.

A chatbot answers a question, and you do the rest. With Agentivity you build specialists with a role, you organise them in teams, and the work comes back finished. A person is asked only when a decision is theirs.

Work with several steps that needs judgement on the way: read documents and file them, prepare and write a note, compare options, handle a request from start to finish. Four examples are ready to import. See What you can do with Agentivity.

No. It takes the repetitive work and comes back to your people for the decisions. See How a person stays in control.

No, to build. You describe what you need to the Assistant, or you build in the Studio with names, descriptions in plain words, and drag and drop.

Two things need a technical colleague: installing the Community edition on a server, and building an application of your own on top of it.

If you can copy a command into a terminal, yes: the install is one command, once Docker is on your machine. See Install Agentivity with Docker. If that is not for you, a hosted workspace has nothing to install: see Start a hosted workspace.

The Studio is in English. Your agents can work in any language that your AI model handles: write their instructions in the language you want them to use.

A computer or a server that runs Docker, and an account with an AI model provider.

Yes. It lasts 14 days, needs no card, and has every feature of the plan you choose, with up to 200 executions. See Start a hosted workspace. The prices are on the pricing page.

The workspace says why: no plan in force, an update under way, or a limit of the plan reached. See Why did my execution not start?.

Yes. It is free of charge, with no time limit. See What the Community edition includes.

Yes, separately. You connect your own account with the provider you choose, and that provider bills you directly for what your agents use. Agentivity adds nothing to it.

Models from Anthropic, OpenAI, Google, xAI and DeepSeek, and Claude models through AWS Bedrock. You can connect several and choose which one each agent uses. See Connect an AI model.

It depends on your provider and on the model. As a rule: each agent that works calls the model at least once, so a team with many members costs more for one request than a single agent. Start small, and set a spending limit with your provider if it offers one.

With the Community edition, on the machine where you installed Agentivity, in one Docker volume. With a hosted workspace, in your workspace, which Agentivity runs, in the European Union; the owner can export everything in one file from Administration > Data. See Update Agentivity and back up your data.

With the Community edition, no. Everything you build stays on your server: agents, teams, workflows, tables and credentials are not sent to Agentivity. The image sends no telemetry, unless you give it the address of a collector of your own. With a hosted workspace, your content is stored in your workspace and processed only to provide the service, as your processor: see the Data Processing Agreement. See Telemetry.

What an agent sends when it works: its instructions, the request, and the documents or rows it uses for that request. It goes to the provider you connected, under that provider’s terms. Do not give an agent what must not leave your server.

Your agents need it to reach the AI model. The Studio needs it to show the ready-made solutions of the marketplace. Building and running workflows that have no AI step and no outside service work without it.

• Questions about building and running
• Questions about the licence, security and editions