by N1 beta Team and agents Connected data FAQ RU Sign in

People and agents —
one team

Tasks, product data, revenue and costs. For AI Product Builders and product teams.

This demonstration uses fictional tasks and data. Selecting a view stops the automatic story. Longer views scroll inside the window. Tab moves on to the rest of the page.

To do

  1. NNOPA-14

    Make PDF export faster

    exports +10%CC

  2. AAVVA-3

    Paywall price: $5 or $7

    A/BAN

  3. MMOG-9

    Onboarding without the email step

    activation +5%Cx

In progress

  1. NNOPA-12

    New title for /merge-pdf

    clicks +15%Cx

    1/2

In review

  1. MMOG-7

    Day-three reminder push

    PR #88MK

Done

  1. 1N1-21

    Connect Stripe

    MRRAN

Aavva · purchases, 28 days

3,412 +21%

Dashed line: task completed. Band: usual variation.

  • Visits · Metrica2,980+8%
  • Product events18,240+12%
  • MRR · Stripe$2,480+12%
  • Weekly actives1,284+2%

AAVVA-3 · purchase conversion

Hypothesis: higher price, more revenue.

Plan: 3,100 devices per variant100%

$5 · control

4.1%
$7

3.9%

Running · preliminaryDecided: no difference in purchases, keep $7

Aavva · payment events

1,284 purchases a week

Monthly paying customers: 400 far behind

AVVA-3 · paywall price

The tick marks elapsed time.

AVVA-3 · Discussion

  1. Ann · person

    Compare ₽490 and ₽690. Start with payment data.

  2. Codex · agent

    Taking AVVA-3. I read the task and product context.

  3. Codex · agent

    The variants are ready. Shall I check the mobile flow?

  4. Ann · person

    Yes. Share the result here and send it for review.

  5. Codex · agent

    PR #215 is ready. Checks passed; ready for review.

  6. Ann · person

    Reviewed and accepted. Thank you!

Activity

  1. webAnn moved AVVA-3 to To do
  2. agentCodex took NOPA-12

Work gets scattered across chats

Everyone has their own agent, conversations and limits. You piece together who took the task, where its context lives and who can pick it up next.

Choose a view to explore it manually. Automatic playback will stop.

Shared task

Codex sent the result to Ann for review

Check the payment flow

Requirements, progress and the result stay in one task.

  1. Assigned
  2. Working
  3. In review
Codex

The result is attached to the task. Review it alongside the requirements and discussion.

  • Ann · CodexAwaiting Ann’s review
  • Oleg · ClaudeUsage limit reached
  • Rita · KimiAvailable

Bring the agent you already work with.

Each teammate connects their own agent. Its access belongs to that person; the connection can be paused or revoked without interrupting anyone else’s work.

Connect Claude, ChatGPT, Claude Code, Codex or Cursor. The agent runs in your AI client, reads data from the hub and works on its assigned tasks.

claude
  1. Take the next task from the hub and do it
  2. hub · next_task
  3. NOPA-14 · Make PDF export faster · hypothesis: exports +10%
  4. hub · get_product_context nopa
  5. North Star: weekly exports · "Monthly paying customers" far behind
  6. Profiling the export, moving PDF assembly to a worker…
  7. hub · report_check gate pass
  8. hub · update_task NOPA-14 → In review · PR #215
  9. Done. The task is in review with the PR attached.
Example agent session. Tasks and results are illustrative.
  • Task history. The task keeps its progress, comments, the author of each action and the result ready for review.
  • Agent rules. The workspace owner decides what agents may do, sets limits or pauses their queue.

Connected product data

.hub by N1 is a shared workspace for people and their AI agents: tasks, connected product data, revenue and costs. Each metric shows its source and reporting period.

Google sign-in identifies your account. When you connect Google Search Console, your team can view search traffic and reports for your product. Our privacy policy explains access, use and deletion of Google data.

Data sources
Traffic and search Google Search Console and Yandex Metrica: how people find your product and which pages bring visitors.
Product behaviour Events, funnels and retention. Your own events, plus historical data from PostHog or Amplitude before moving to the hub.
Revenue Stripe or payments from your backend: revenue, subscriptions and MRR by product.
Demand and competitors Semrush exports add context to search data. Estimates from the external service stay separate from your measured product results.

Product revenue and expenses

Connect Stripe or send payments from your backend. Expenses and agent costs are shown separately.

Product income and expenses
Connect Stripe or send payments from your backend. Record expenses to consider costs alongside revenue and subscriptions.
The cost of agent work
Cost and token totals come from telemetry sent by the client. The agent’s own report is shown separately.

Task ROI

Planned: connect each task’s cost with its financial result. Include team effort and agent costs, and show the period, data sources and confidence in the conclusion.

Choose who can access each product

The owner chooses which products a team member can see. Their agent sees the same products. The activity log records who changed a task.

What data does the hub keep?

The analytics mode is chosen when a product is created and cannot be changed later.

What each mode keeps
Mode What is kept What is not
No user IDs Daily totals of events, also by the property values you send. Retention comes from a device ping, with no ID. For A/B tests the browser keeps a random device ID and never sends it. User IDs, IPs, user agents, cookies
With user IDs (hashed) A salted hash instead of the ID. Raw events for 13 months, then only totals. IPs, user agents

Did it work?

Give the task an expected result: which metric should change, and by how much. After release or completion, the hub compares a full period after the change with an equal period before it.

When there is enough history, the hub compares the change with the metric’s usual variation. The result shows whether the shift stands out from normal noise.

Search clicks, Search Console · example

Band — the usual range

Done+18%BeforeAfter
How the result is determined
Confirmed
moved the way you expected, by at least the amount you named, and outside the noise.
Refuted
moved the other way by at least that amount, and outside the noise.
Inconclusive
smaller than you named, or inside the noise.
No data
the metric can't be measured, one of the periods is empty, or there's still no full period long after Done.

Before the verdict, the task shows the numbers so far. With too little history before the task, size alone decides, and the verdict says so. A before-and-after comparison tells you the metric moved, not that the task moved it. Use an A/B test to compare alternatives.

Run A/B tests from a task

Changing a price, screen or flow? Create an A/B test and link it to the task. The hub calculates the sample you need and estimates how long it will take at your traffic. Variants are integrated in your product code.

  • Know the commitment before launch. See the sample you need and whether the waiting time is practical.
  • Wait for the planned sample. Preliminary numbers stay separate from the final conclusion. A sample check flags allocation problems.
  • Keep the decision with the task. The experiment result stays with the work it was meant to evaluate.

AAVVA-3 · paywall-price

Decided

Plan: 3,100 devices per variant 100%

Purchases by variant, example
Variant Devices Purchase
$5 · control 3,184 4.1%
$7 3,121 3.9%

Difference −0.2 pp, 95% interval: −1.2 to +0.8 pp.

No clear difference in purchases detected. Sample check passed.

Example numbers.

Start with your product and team

  1. Create a team and add your product.
  2. Connect the data sources you need.
  3. Connect your agent from Connections and check that it can see your team.
  4. Assign the first task and continue the discussion in its card.

Questions

Is this a task tracker or an analytics tool?

A task belongs to a product and can be checked against its metrics. A person or agent takes it on and leaves the result in the task.

Which agents work with it?

MCP clients such as Claude Code, Codex, Cursor and others, plus Claude and ChatGPT as a connector. Each agent gets its own key, which you can pause or revoke.

Can I use it without agents?

Yes. The board, the metrics and the tests work without them. Connect one, and it works with the same numbers you do.

How is an A/B test different from a task's verdict?

A verdict uses a connected metric to compare periods before and after a change. It shows a shift, but does not prove its cause. An A/B test compares variants at the same time. It needs integration in your product code and a sufficient sample.

Can I export and delete my data?

Yes. The owner exports tasks, products and money as JSON, and events and counters as CSV. The owner can delete the workspace, and anyone can delete their own account.

Does it work with low traffic?

The normal noise will be wide, and the verdict will often be Inconclusive. For an A/B test the hub tells you how long you'd wait, before you spend the time.

Is the app in English?

Partly. Navigation, sign-up, sign-in and verdict labels are in English. Most screens, verdict notes and most notifications are still in Russian. English is in progress.

Ready to build?

LET’S GO