Noots captures the meeting. The board updates itself.
Send the bot into the call, or capture from your own browser. The tasks, dates and owners land on the right board — and an AI project manager keeps them moving, on a leash you set.
Free for 10 projects and 10 people. No card.
We can't start the fit-out until the switch order lands.
I'll chase the supplier today and confirm the date.
Then the survey slips a week. Let's re-plan around it.
Fit-out waits on the switch order
Chase the supplier — Jo, today
Site survey slips one week
Joins the calls you already run. Imports the boards you already have.
Three jobs off your desk — and the workspace that runs the rest.
These three are what most teams come for. They arrive on a workspace that also runs the boards, the chat, the calls, the calendar, the mail and the CRM — pick a tab below; every one is a screen the product really ships.
From meetings to board, instantly.
Send the Noot bot into the call, or capture from your own browser if your team will not have a bot in the room. Either way the project updates itself.
- Speaker-labelled transcript with timestamps and talk share
- Decisions, action items, risks and topics pulled out by name
- Share a screen and the bot can follow that too — what was demoed, what broke, what got decided. Business plan, off until you turn it on, and the frames are never stored
- Tasks land on the right board, dated, assigned and linked back to the minute they came from
- Zoom, Teams, Meet and Webex — or no bot at all
We can't start the fit-out until the switch order lands.
I'll chase the supplier today and confirm the date.
Then the survey slips a week. Let's re-plan around it.
Fit-out waits on the switch order
Chase the supplier — Jo, today
Site survey slips one week
An AI project manager, on a leash you set.
Two dials, not one. How much it may do without asking, and how far ahead it is answerable for. It reports what it deliberately left alone.
- Turns a brief into phases, epics, tasks and checklists — after interviewing you in five questions
- Assigns capacity-aware across every project a person is on, and never to somebody who is away that day
- Reads “about 3 days behind” as a re-date, ripples the dependent dates, tells the owners
- Answers “will we hit the date”, and the same question about the budget — escalating the ones arithmetic says cannot hold
- Books the weekly, the per-team check-in and every milestone review; writes each agenda, invites the people, makes the room and sends the bot
- Escalates blocked work to a named human — and every change it makes is individually reversible
Site survey → Mia (68% loaded, lightest of three)
Undo“about 3 days behind” → moved 4 tasks
UndoSwitch order blocked → raised to the team lead
UndoFit-out crew has no lead — suggest Ravi
Needs you11 tasks due past the milestone were left alone. Every change above is individually reversible.
Chat wired to the work, on one clock.
Every project gets a channel. Every person keeps their own timezone and their own working hours, and the scheduler respects both.
- Find a time that resolves each person in their own zone — and blocks on approved leave
- An event carries who is coming, what it is about and its agenda — wherever you create it
- One rail: your people under their own heading, rooms under theirs, and one search that starts in the room you are reading
- Voice and video built in, with live notes while the call is still running
- Warns you before you message someone at 21:40 their time, and offers to schedule it
- Summarise a thread into the project it belongs to
Jo Supplier confirmed — switches land Tuesday.
Alex /summarize
Send now, or schedule for their morning?
14:00 UTC works for all three. Mia is on approved leave Friday.
The parts nobody misses.
Notes that write themselves, dates that move on their own, and an hour on Thursday that nobody had to negotiate.
Nobody has typed up a meeting since March. The actions are already on the board with names on them by the time we walk out.
Took me a month before I'd let it change anything by itself. Two of my four still only propose, and they'll probably stay there. The other two re-date themselves on a Friday and I've stopped arguing with it.
Alert landed at 07:41 and the ticket was open with the clock running before I'd got my coat off. Same alert again at 08:03 and it didn't open a second one, which is the bit I actually care about.
It arrives already knowing how your trade works.
Pick what you do when you create the workspace. The projects, the tabs on every board and the tools that come switched on are different for a service desk than for a site office — not a different colour scheme, a different shape.
IT & Managed Services
SLA countdowns on every ticket, hours that turn into invoices, and monitoring alerts that open the ticket themselves.
07:40 — an alert fires before anyone is at a desk
PagerDuty, Datadog, Zabbix, Alertmanager, UptimeRobot or any JSON webhook posts to a Noots endpoint. The vendor adapter turns the payload into a task on the right board with the right fields, attaches an SLA, opens an incident channel and pages the on-call team. A repeat of the same alert updates the ticket it already made instead of opening a second one.
- Service Desk
- Client Projects
- Internal IT
Six tools, one workspace.
On the left, six subscriptions a team of forty usually pays for separately. On the right, what Noots already ships in their place. Keep the ones you love — everything here imports from them.
- A board tool
- Boards, list, timeline, Gantt and capacity
- A meeting recorder
- The bot, the transcript and the analysis
- A chat app with calls
- Channels, DMs, voice and video
- A scheduling tool
- Find a time in everyone's own timezone
- A starter CRM
- Contacts and deals tied to the boards
- An automation builder
- Triggers, conditions and actions per project
Every one of these, and Autopilot does them.
- Breaking a brief into phases, tasks and checklists
- Working out who has room for it, across every project they are on
- Chasing the status update, and re-dating what slipped
- Rippling the dates that depended on it, and telling those owners
- Answering whether the delivery date and the budget still hold
- Booking the week's meetings, writing each agenda and sending the bot
- Finding the person who can unblock it, and escalating to them
- Writing the summary at the end of the week
It runs on a schedule you set, at the level of autonomy you choose, and every change it makes can be undone one at a time. Autopilot is on the Business plan.
Run the numbers against one hire.
One subscription against the stack it replaces, plus the project manager Autopilot does the weekly work of. Open the breakdown to tick what you actually pay for, and put your own salary figure in — we are not going to guess either.
Autopilot is included on Business — it is not an add-on. Everything else on that list is a surface Noots already ships, so nothing here is a plan to build one.
$103,400 a year
What that stack figure is made of3 of 5 tools ticked · monday.com · Fireflies.ai · Slack · US list prices read 31 July 2026
Other tools' prices are US list, read from each vendor's own pricing page on 31 July 2026, at the annual-billing rate, on whichever plan that vendor badges as its recommended or most-popular tier. Zapier is priced per task rather than per person, so its line does not scale with your headcount.
Autopilot is not a person and this is not a like-for-like swap — it plans, assigns, chases, re-dates and escalates, and it does not sit in the room with a client. The comparison is what it costs to have that work done.
The whole company, in one glance.
Because every meeting is captured and every board is live, leadership gets a real window instead of a reporting ritual.
- No status decks — the status is the board
- No sitting in on calls to know what was decided
- No waiting until Friday for the write-up
Everyone sees only what they are allowed to see. A private project stays out of every roll-up, including this one.
Beckett Street fit-out
Switch order blocked — lead notified
Q3 platform migration
Ahead of the milestone by 2 days
Northgate rollout
Survey re-dated after this morning's call
Encrypted, permissioned, and audited.
Every control below is built and running today.
Encrypted in transit, and credentials sealed at rest
TLS on every connection, and every third-party credential we hold is stored under an AES-256-GCM envelope.
SSO, SCIM, and two-factor you can require
Bring your own identity provider, sync joiners and leavers, and require a second factor of everyone — refused at the request, not at the page.
Role-based access, enforced in SQL
Private projects are filtered in the query, not hidden in the interface.
Audit log and IP allowlisting
Every administrative action is recorded; access can be pinned to your network.
Run it on your own model. Nothing leaves your server.
Point a workspace at an OpenAI-compatible server you operate and every AI call goes there instead — meeting summaries, mail triage, chat, Autopilot, your AI teammates and Hoot. Not a copy. Not a fallback.
The requirement is a protocol, not a product, so there is no compatibility list to go out of date: if your server answers an OpenAI-compatible chat completion, this works — vLLM, Ollama, LM Studio, LocalAI, llama.cpp, TGI, LiteLLM, Together, Groq and Azure OpenAI all do.
What it means, and what it doesn't.
- Noots never stores your meeting audio. There is no file storage in the product at all — audio is transcribed in flight and the recording is not kept.
- Transcripts and summaries are stored as ordinary text in our database. Private AI decides which model reads your content; it is not database-level encryption, and we do not claim any.
- Business plan, owner or admin only — re-checked on the server every save, never inferred from the screen.
- Every call is logged where you can see it: when, which feature, which host, and whether it worked. Never what was in it.
- One exception, and it is one you make: if you separately connect a third-party AI provider with your own key, work you push to it goes there.
One address, re-checked every call
A base URL, a model and a key. The key is sealed with AES-256-GCM and never shown again. The address must be https and publicly resolvable, and Noots re-validates it, pins it and refuses redirects on every call — a hostname can be repointed after you save it.
A killswitch that is on by default
If your server is unreachable, the feature that needed it fails and names your server. Nothing is retried on ours. Allowing a fallback is a separate switch you have to find and read.
A test that audits your server
Five checks before anything routes — address, certificate, which model answers, a real completion you can read, and the same request with no API key, which is how most self-hosted servers turn out to be open to anyone with the URL. It re-runs daily and only ever records a failure.
Your transcription too
Point the audio path at an OpenAI-compatible transcription endpoint and recordings are transcribed on your hardware. Two honest edges: that API has no diarization, so those meetings come back without speaker labels, and a third-party meeting bot arrives as a link, so it fails rather than quietly going elsewhere.
Start with your next meeting. The board catches up.
Create a workspace, tell it what you do, and put the bot in your next call. Nothing to migrate first.
Free for 10 projects and 10 people. No card.