← All docs

Outcome Deals

Your workspace

Everything about your deal lives in one place in the browser. This page is a map of it: where to click, what each tab holds, and what happens inside a chat.

Getting there

Your outcome opens the first time you click Set up the outcome on your deal page. After that, sign in at app.nimroboai.com and choose Outcomes in the sidebar, or use the Open outcome link on your deal page.

There is nothing to install. The Nimrobo Mac app is a separate product for running your own outcome loops, and your deal doesn't use it.

The top of the page

GGoogle organic search clicks
Chat
Run next action
Target50 → 100Window24 days leftCurrentView receipt →
OutcomeLeversHypothesesExperimentsSessionsFilesSettings

The arrow ▾ opens

SetupDefine the north star, levers, and guardrails.
RunTake one action on a lever and measure it.
UpdateReview the evidence and decide what to try next.
A schematic of the top of an outcome. Example data. The real screen differs in detail.
What it does
1. ChatA free conversation with the agent about this outcome.
2. Run next actionStarts a Run. The arrow opens Setup, Run, and Update.
3. Deal bannerYour target, the days left in the window, and a link to the deal.
4. TabsWhat the agent has planned, tried, and measured.

Run next action starts a Run, the action you'll use most. The arrow beside it opens all three actions: Setup, Run, and Update. What each one does.

On a new outcome, start with Setup

A new outcome has nothing planned yet, so a Run has nothing to act on. Start with Setup. After that, Run next action is the button you'll come back to.

Chat opens a conversation with no action attached. Use it to ask why the agent chose a change, what a result means, or what it plans next.

The tabs

Most tabs are the agent's working notes, kept in the open so you can see why it does what it does. You don't need to manage them.

TabWhat it showsWritten by
OutcomeYour north star, the number your deal is measured on, with its trend. Whether a new reading is due. Your last five chats.The agent records readings
LeversThe kinds of change the agent believes can move your clicks: sharper titles and descriptions, stronger pages, new pages. Each stays a candidate until evidence proves or retires it.The agent
HypothesesPredictions, each written before a change ships: what will change, and how you'd know it didn't work.The agent
ExperimentsHow each prediction is being tested: the changes made under it, their latest readings, and the verdict once there is one.The agent
SessionsEvery chat and run, newest first. Open one to read it or pick it back up.You and the agent
FilesThe outcome's folder: its goal, its operating rules, the checks a change must pass, a copy of your deal terms, and a folder for each run. Read-only.The agent
SettingsThe outcome's name, its Search Console connection, and which outside data the agent may spend credits on.You

Inside a chat

Setup, Run, Update, and Chat all open a chat. The agent works in the open: you see its steps as it takes them, and the chat stays in Sessions afterward.

When the agent needs you

Sometimes the agent stops and asks. The question appears as a card in the chat:

CardWhat it's askingHow to answer
The agent is askingA question it needs answered to continue, like which page to leave alone.Type an answer or pick an option.
External data budgetPermission to spend data-credits on outside data, like a live look at Google's results for a search.Allow once, Allow this run, or Deny.
Sandbox approvalPermission to run a step. You'll see these when the chat is in Manual mode.Allow once, Allow this run, or Deny.

A card waits 10 minutes. If nobody answers, it expires and the run carries on without your answer. A permission request that expires counts as denied. A yellow dot next to an outcome or a chat in the sidebar means it's waiting on you.

Leaving a run

You can close the tab while a run is working. It keeps going, and you'll find it in Sessions when you come back. A green pulsing dot in the sidebar means it's still running.

Auto and Manual

The pill beside the message box sets how much the agent asks before acting in this chat:

  • Auto, the default: the agent's tools run without asking.
  • Manual: every tool except reading asks first.

Either way, the agent can't publish to your site. Its tools work inside your outcome's own workspace, and changes reach your site only through you.

Settings worth knowing

  • Your Search Console connection shows Bound to this deal. Your deal is measured through it, so it stays attached until the deal settles. Keep it connected under Connectors in your workspace.
  • External Data lists the outside data sources the agent may use, with a per-run credit ceiling.
  • The number your deal is measured on can't be changed. It's a term of your deal.