This plugin adds a conversational AI agent to your Bubble app, one that can actually take action rather than just chat. You connect your own Backend Workflows as tools, and the agent decides on its own when to call them to answer a question or carry out a task. It might look something up in your database, run a calculation, send an email, or create a record - whatever your workflows can do - and then it replies in plain English using whatever those workflows returned.
The agent runs a full tool-calling loop: it reads the conversation, decides whether it needs a tool, calls it, reads the result, and then either calls another tool or writes its final answer. All of that happens inside a single action. You send a message, the agent handles the rest.

Tools are just your existing Backend Workflows, so there's nothing new to learn to create them. You don't have to describe them by hand either - the plugin scans your app, finds every API workflow whose name ends in a suffix you choose (the default is _agent), and builds the tool list for you. You then add plain-English descriptions so the agent knows when to use each one.

It works with OpenAI, Anthropic, or Google Gemini using your own API key, and you can switch provider or model at any time without touching your workflows.

The conversation comes back as a fully typed list of turns that drops straight into a repeating group. Each turn carries the message, the reply, any tool calls that were made, token counts, timing, and success or error details, so there's no JSON parsing needed to display a chat.

Features included:

- Multi-provider support: OpenAI, Anthropic, and Google Gemini.
- Automatic tool discovery from your Backend Workflows, filtered by a naming convention you control.
- Streaming answers for a live typewriter effect on OpenAI, with non-streaming answers on Anthropic and Gemini.
- Extended thinking / reasoning mode on models that support it.
- A configurable safety cap on the tool-calling loop, plus automatic retries when a tool call fails.
- Two authentication modes for tool calls: as the signed-in user, so cookies and privacy rules apply, or as admin using an API token.
- A server-side action, so the agent can run inside Backend Workflows, on a schedule, or from a webhook with no page involved.
- Stop a response mid-stream, clear the conversation, or reload a saved conversation from your database.
- Token usage reported both per turn and across the whole session.
- Configurable system prompt, maximum turns, retry count, and tool choice.
- Debug mode with detailed browser and server logs for troubleshooting.

Six actions in total: Send Message and Send Message (Backend) run the agent, Generate Tools builds your tool list, and Stop Generation, Clear Conversation, and Load Conversation History manage the conversation.

A note on complexity: the chat side of this plugin is quick to set up, but the tools side does ask a little of you. You'll need to be comfortable creating Backend Workflows and exposing them as public API workflows, since those are what the agent calls. If your workflows live on a different app, or need to bypass privacy rules, a basic understanding of Bubble's API tokens helps too. None of it is difficult, but it's worth knowing before you start.

You'll need your own OpenAI, Anthropic, or Google Gemini API key.

A full demo, setup instructions and support is included.