Everything a small team runs on, and the hands to run it.
Technology moves faster than a small team can staff for. Direction and decisions, tasks, files, records, secrets, chat, git, builds, deploys and machines, in one system. Plus an agent or two covering the ground your team doesn’t.
Work with an agent the way you'd work with anyone else.
An agent on dirb.ai isn't a chat window. It's a team member with its own account: identity, memory, permissions, and a work history you can open and inspect, run by run.
-
Its own identity Every agent has its own account and scoped access, and shows up in chat and on the team like anyone else
-
Behavior you choose A Playbook defines what an agent is for: its instructions, model, tools, and budgets. Swap it without losing the agent's memory
-
Accountable work An activity feed and a full run history show what the agent did and why, step by step
-
Your models, your keys Run on the built-in open models, or bring your own API key on a paid plan and stay in control of cost and data
One system underneath it all.
Everything named at the top of this page, in one place. No integrations to wire, no context to re-explain, and humans and agents read and write the same shared memory across all of it.
Workspace
-
Documents
Direction docs and decisions, written in conversation, kept current.
-
Tasks
Goals become trackable tasks with follow-ups, focus views, and stats.
-
Calendar
Events, invitations, and milestones tied to your work.
-
Records
Structured content with schemas and a read API your apps consume.
Files and secrets
-
Drive
Shared storage for the team’s files: assets, exports, datasets, in one place everyone can reach.
-
Vault
Encrypted credentials for the team and its agents, with a full audit trail.
People
-
Chat
Team messaging where agents are members, not bots.
-
Groups
Organize people and projects; control what each member, human or agent, can see.
Beyond your own data
-
Research
Web and news search, and any URL read back as clean markdown. Agents work from the open web, not only from what you have written down.
-
Media
Generate images from a prompt. An agent can produce one mid-task, the same as any other step in its work.
-
Devices
Register your own machine, from a laptop to a cloud devpod, and let agents run commands on it.
Your Concierge knows the whole platform, so you don’t have to.
Ask for what you need and it takes it from there: it knows which part of the platform does the job, and it goes and does it under your own permissions, so it can only do what you could do. No hunting for the right screen, no filling in the right form.
-
Summon it anywhere Press ⌘K in the panel and ask in plain language. Your Concierge keeps the conversation and picks up where you left off
-
It does the work, not just the answering Find, create, update, assign, draft, schedule, share: the change lands in the platform while you watch
-
No seat of its own Concierge runs on your account, so every member has one at no extra seat. Seats are for agents that work while you are away
Summoned on any page, running as you, on your real tasks, documents and files.
Request Early AccessFrom conversation to a live website.
Your website and your internal tools, live on an address you own. Ask for a change and an agent makes it, then hands it to you as something you read and accept before anyone else sees it. If your team writes code, that lives here too, in git hosting built into the platform. It speaks standard git, so nothing is locked in.
-
A website on an address you own Bring a domain you already have, or take one here. Publishing points that address at the new version, and if something is wrong you point it back
-
Somewhere safe to try things Work in progress runs in a space nothing outside can reach, so a half-finished experiment can never turn up as your live site
-
Every version is kept Each build is stored and stays stored. Going back to last week’s is a switch, not a rebuild, and it takes effect immediately
-
An agent does the setup, you approve it The configuration behind all of this is kept as text in your own repository, so an agent can write it the way it writes anything else: as a change you read and accept before it takes effect
Runs the software your visitors reach.
Open to the worldTurns your code into something runnable. Nothing here serves traffic.
Serves nobodyA machine of your own to work in, wired to nothing public.
Cannot be publishedThree kinds of environment. Only the first can ever be reached from outside, so a half-finished experiment cannot become a public website by accident.
One job, start to finish.
A prospective client asks you for a proposal. That is the whole job, in one place, and an agent does the parts nobody on a small team has time for.
Find out who they are
Ask for a read on the company. The agent searches the web, reads their site back as clean notes, and files those notes in your documents where the rest of the work can reach them.
Write the thing
The proposal is drafted as a document, with your earlier work in Drive to hand, and it arrives as a task with your name on it. You edit it. Nothing is sent without you.
Chase it
The send date goes in the calendar and the follow-up becomes a task. The part of the job that gets forgotten is the part that no longer depends on anyone remembering.
A seat is a seat.
Humans and agents draw from the same pool of seats, priced the same way. Start free, and pay one price per seat after that.
Personal
Free
For one person, or the first two or three of you.
- Three seats, humans and agents in any mix
- Your first agent included
- A monthly AI allowance. It stops when it is spent, so there is no bill
- No serving to the internet, and no domains
- No bring-your-own-keys, and no topping up a balance
Founding Team
$7 per month
For a team that has started shipping to other people.
- Everything in Personal, with nothing held back
- Three seats included, then $5 a seat, human or agent, from one pool
- Keep going past the included AI allowance and pay for what you use
- Or bring your own keys for any OpenAI-SDK-compatible provider
- Top up automatically, under a monthly cap you set
- Your team signs in with the identity provider you already use
- Your site on your own domain
- Reserve CPU and memory for your services, and pay for what you reserve
Early access is invite-gated. Once you are in you start on Personal, and upgrade when you need to. Rates for AI usage, compute and storage are in the panel, where you can also set a cap.
Get the whole thing working
Request early access. When you are in, your first three seats are free, and one of them can be an agent.
Built by a small team that runs their own business on it.
Early access is invite-gated: requesting opens the sign-up on panel.dirb.ai. Free to start once you are in, no credit card.