---
title: "What is the agent?"
description: "What the Rank Prompt agent is, what it can read and do for your brand, who can use it, and how it is paid for."
url: https://help.rankprompt.com/en/articles/70-what-is-the-agent
collection: "Agent"
updated: 2026-11-20
language: en
---

# What is the agent?

What the Rank Prompt agent is, what it can read and do for your brand, who can use it, and how it is paid for.

The agent is a chat built into Rank Prompt. It reads your brand's data in the app (reports, prompts, competitors, sources, tasks, articles, audits) and acts on it: it explains what changed, finds what to fix, and does the work, from creating tasks and running reports to drafting articles.

You do not have to copy data into it or explain your situation. Ask from any page and it starts from what is on your screen.

The agent is on every plan, the Free plan included.

* * *

## What it can do

-   **Explain your results**: your Visibility score, the prompts that name your brand and the ones that miss it, the AI platforms, competitors and the sources AI cites
    
-   **Find what to fix**: gaps against competitors, sources worth pitching, prompts to write content for
    
-   **Manage your work**: create, update and close tasks; work through a task end to end and move it to **Done** when the deliverable is ready. It can also hand a task off to the part of Rank Prompt that does that work, the same as the task's button in **Tasks**, and tell you where the proposal waits for your approval
    
-   **Run and change reports**: run a report, create a report that runs once or on a schedule, change a report's schedule, add or edit prompts
    
-   **Write content**: draft articles, edit them, and publish them on the plans that include content
    
-   **Run outreach** (Pro): write pitches, send an email from your connected Gmail or Outlook or launch a campaign
    
-   **Research**: keyword research, where your tracked keywords rank on Google, and a web search when **Web search** is on in Agent settings
    

Each capability follows the plan of the feature behind it. When you ask for something your plan does not include, the agent tells you and shows the plan that does.

### It asks before it spends or changes something outside

Anything that costs credits, deletes something, or writes outside Rank Prompt (publishing an article or sending an email, for example) shows an **Action required** card with the cost. It only runs when you click **Confirm**. Creating a task, saving a memory or writing a draft does not ask.

A job of several steps comes as one plan card instead: every step with its cost and the most credits the whole plan can spend. **Approve plan** runs all of it without asking again, **Confirm each step** brings back a card before each step that needs one, and **Cancel** stops it. A plan made for an earlier message expires; ask again for a new one.

### It connects what a job needs

When a job needs a connection you have not made (Google Analytics or Search Console, Gmail or Outlook, Google Business Profile, GitHub or your publishing site), the agent shows a card with **Connect**, or **Reconnect** when a connection stopped working. The sign in, or your site's form, opens right there. Then ask again.

### What it cannot do

-   Change your plan or buy credits (it can show you the plan that fits)
    
-   Post a reply in Community (it drafts the reply for you)
    
-   See brands you cannot open yourself
    

* * *

## Skills

A skill is a job the agent knows well: it reads the right data, does the work and ends on one next step. Examples: **This week's visibility**, **What holds my visibility back?**, **Me vs a competitor**, **Sites AI cites that skip me**, **Plan a month of articles**, **What should I do this week?**.

Open the **Skills** page (the sparkles button in the agent's side panel) to see them all, in the tabs **Visibility**, **Citations**, **Outreach**, **Content**, **SEO** and **Tasks**. In Agency mode on **Agency Plus** it shows the **Agency** skills instead, such as **Portfolio briefing** and **Deliverable draft**. Each tile shows what it needs (Google Analytics or Search Console, a connected CMS, a Business Profile, GitHub) and the most credits it can spend; a skill your plan does not include shows the plan it needs. A skill runs in a new conversation in the side dock.

**Ask agent** in a page header runs that page's skill, for example **Explain my AI visibility** on the AI visibility page. The same **Ask agent** button sits wherever the app offers the agent, on a task, a Home card or a row, and asks about that thing.

* * *

## In your own AI tool

The Rank Prompt MCP server gives Claude Code, Cursor, ChatGPT and any other MCP client the agent's tools and the same skills, so you can ask from the AI tool you already use. There the calls spend your credits and your API request allowance, not your AI budget, and Free accounts can connect too. See [Connect Claude Code](https://help.rankprompt.com/en/articles/98-connect-claude-code.md) and [Connect Claude (claude.ai and Claude Desktop)](https://help.rankprompt.com/en/articles/134-connect-claude-claudeai-and-claude-desktop.md).

* * *

## Who can use it

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Seat</p></th><th colspan="1" rowspan="1"><p>The agent</p></th></tr><tr><td colspan="1" rowspan="1"><p>Owner, editor</p></td><td colspan="1" rowspan="1"><p>Full use</p></td></tr><tr><td colspan="1" rowspan="1"><p>Team viewer</p></td><td colspan="1" rowspan="1"><p>No chat</p></td></tr><tr><td colspan="1" rowspan="1"><p>Client seat (Agency Plus)</p></td><td colspan="1" rowspan="1"><p>Only when the agency turns on <strong>Your clients can use the agent</strong>, and read only</p></td></tr></tbody></table>

On **Agency Plus**, **All brands** on the Agent page (or **Agency mode** in the brand switcher) lets you ask across every brand in your portfolio: which clients slipped this week, which one needs attention first.

* * *

## AI budget

The agent spends **AI budget**, a dollar balance separate from your credits. AI edits of text and images in the article editor spend it too. A task you hand off is the exception: it costs credits at the price of the part that does the work, never your AI budget, and the confirmation shows the cost first.

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p>Plan</p></th><th colspan="1" rowspan="1"><p>AI budget a month</p></th></tr><tr><td colspan="1" rowspan="1"><p>Free</p></td><td colspan="1" rowspan="1"><p>$1</p></td></tr><tr><td colspan="1" rowspan="1"><p>Starter</p></td><td colspan="1" rowspan="1"><p>$5</p></td></tr><tr><td colspan="1" rowspan="1"><p>Pro</p></td><td colspan="1" rowspan="1"><p>$10</p></td></tr><tr><td colspan="1" rowspan="1"><p>Agency</p></td><td colspan="1" rowspan="1"><p>$15</p></td></tr><tr><td colspan="1" rowspan="1"><p>Agency Plus</p></td><td colspan="1" rowspan="1"><p>$25</p></td></tr></tbody></table>

The Free plan's budget resets on the 1st of each month; paid plans reset with the billing cycle. When it runs out, convert credits into more on **Plans > AI budget**, or run the agent on your own OpenAI key. Inside a brand, a teammate's chats spend the brand owner's budget, the way credits do. See [Agent settings and your own OpenAI key](https://help.rankprompt.com/en/articles/71-agent-settings-and-your-own-openai-key.md).

### What an answer costs

-   **The help center search**: before it answers most messages, the agent looks up this help center so it explains the app as it works today. That search is part of the answer's cost, and **AI usage** on **Subscription > Usage** lists it next to the answer. With your own OpenAI key active, it runs on your key too
    
-   **A few cents to start**: the agent starts a reply only while the AI budget left covers a long answer, a few cents. Below that it says the budget ran out, even if the balance is not exactly zero
    

### Very long conversations

A conversation can only grow so long. When it gets close, the chat says how much of its context is used and offers **New chat**. Past the limit, the conversation ends: it takes no more messages, says it is too long to continue and offers **New chat**. Everything it did stays: the conversation is still in your history, and the tasks, reports and drafts it made are saved. Start a new chat to keep going.

* * *

## Next steps

-   [Using the agent](https://help.rankprompt.com/en/articles/73-using-the-agent.md): open it, what to ask, commands and mentions
    
-   [Agent memory](https://help.rankprompt.com/en/articles/68-agent-memory.md): what it remembers between conversations
    
-   [Agent settings and your own OpenAI key](https://help.rankprompt.com/en/articles/71-agent-settings-and-your-own-openai-key.md)
    

* * *

## Questions?

Email [support@rankprompt.com](mailto:support@rankprompt.com) and I'll help.
