Assistant
Work with AI right in ctxflows.
The built-in assistant does anything you can do in ctxflows, just by being asked: it opens any page, changes what it shows, creates items, writes docs and starts workflows. It brings the context layer and the workflows to everyone on the project who works without a terminal.
Anything you can do in ctxflows, just by asking
The built-in assistant is an AI assistant for your project, no terminal needed: it has the tools and commands to work all of ctxflows, and does anything you could do on a page when you ask.
It takes you to any page, works the one in front of you, creates and moves items, writes docs, starts or builds workflows and answers actions as you. It works under permissions you set per project, and deleting for good stays off unless you turn it on.
Finds anything in your project
Ask where something is or why it happened, and the assistant reads the runs, docs, records and journal entries that answer it.
It explains what it found in plain words, and it knows the page you're on, so “this item” means the one in front of you.
Pick the model
You choose the model the assistant answers with, from current models of several AI providers.
Switch whenever a question calls for it: a model that reasons deeply for a hard problem, a quicker one for a small change.
Searches the web when it needs to
When a question reaches past your project, the assistant searches the web and reads the pages it finds.
Ask what changed in a framework's latest release, and have it write what matters for your project into a docs page or an item.
For everyone on the project, nothing to install
Everyone on the project has the assistant in the web app, with nothing to install and no coding agent of their own needed.
It opens beside any page and stays open as you move around; on a phone it fills the screen. Each person's chats are their own, and what the assistant changes is in the project's journal for everyone.
A morning check from your phone
A product manager reviews the overnight work without a laptop, a terminal or a coding agent of their own.
“What's waiting on me?”
The assistant lists the actions waiting on you, oldest first: an approval for a docs change and a question from a run.
“Why did the import run fail?”
It reads the run and its journal, and explains where it stopped and why.
“Add a bug for it.”
It files a bug on the triage board from what it found, in Reported, where the triage workflow picks it up.
“Approve the docs change.”
It shows the change and its checks; you confirm, and it answers the approval as you.
No screen hunted for, no terminal opened, and the journal shows each change as the assistant's, acting for you.
Questions about the assistant
Does the AI assistant need a terminal or Claude Code?
Which models can I use?
What can it change?
Does it write code?
Who sees my chats?
Ask your project anything
Open the assistant on any page and say what you need.