> ## Documentation Index
> Fetch the complete documentation index at: https://docs.usequotable.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Quotable Agent

> Quotable Agent is your AI visibility copilot: it answers from your tracking data, brand context, knowledge base, analytics and the web, and acts in your connected apps after you approve.

> [!NOTE]
>
> **Plans:** All plans, [limits](https://docs.usequotable.com/workspace/plans-and-limits.md#how-credits-work) vary · **In the app:** Sidebar → Agent → New chat

**Quotable Agent** is the assistant inside Quotable. Ask it why a competitor overtook you on Perplexity, which sites to pitch, or what changed last week, and it answers from your own data. It can also draft outreach, edit a table, start an agent or act in the apps your brand connected. Open **Quotable Agent** from the sidebar and click **New chat**, or click **Ask Quotable Agent** inside a table.

Try asking

Which prompts did we lose on ChatGPT in the last two weeks, and who was cited instead?

## What Quotable Agent knows

| Source | What Quotable Agent reads |
| - | - |
| **Tracking data** | Visibility history for the last 14 tracked days, the latest result of each prompt per engine, top cited sources (labelled yours, a competitor's or independent), and competitors |
| **Brand context** | Approved brand records plus team and personal context from the [Context Manager](https://docs.usequotable.com/actions/context-manager.md), ranked for each question |
| **Knowledge base** | Verbatim excerpts from your indexed calls, documents, CRM notes and tickets |
| **Analytics** | Google Analytics 4 and Search Console, once [connected](https://docs.usequotable.com/integrations/analytics.md) |
| **The web** | Live web and news search, and pages it is allowed to read |
| **Your attachments** | Files and documents you add to the message |

Quotable Agent grounds every claim in what it read. It does not invent numbers, states the sample size the first time it quotes a rate, calls small samples "directional", and checks the premise of your question against the data before explaining it. Web facts come with links. When it recommends sites to pitch, it suggests only independent sites, never competitor-owned ones.

Activity shows each step as it runs, for example "Reading latest prompt results", "Reading cited sources", "Searching the knowledge base" or "Searching the web". Expand the activity to inspect results, failures or partial changes. Workspace references use readable names; internal record IDs and version hashes are hidden from answers.

## What Quotable Agent can do

> [accordion] Answer with charts and tables
>
> Quotable Agent can draw a visibility score trend, share of voice or engine mentions as a chart in the chat, and shows analytics results as tables.

> [accordion] Research the web
>
> Up to three web or news searches and three page reads per message. Quotable Agent only opens pages you wrote in your message, pages a search returned in the same message, or the home page of your brand or a tracked competitor.

> [accordion] Write a document
>
> Ask for a document, analysis, plan or content brief and Quotable Agent saves it in the [Library](https://docs.usequotable.com/actions/library.md). The document opens in a side panel beside the conversation, where you can read and edit it while you keep chatting. A card in the thread reopens it with **Open**. In the same chat, ask Quotable Agent to revise a document it saved and it edits it in place.

> [accordion] Write an article
>
> Ask Quotable Agent to write an article and it saves the finished draft as an editable document in the [Library](https://docs.usequotable.com/actions/library.md), shared with members who can access the brand. Writing an article uses credits like any other agent work. There is no separate monthly article allowance.

> [accordion] Work in tables
>
> List, read and create tables, edit cells, add columns and add rows. Values are checked against each column's type and options, and every write is read back to confirm it: you see "All changes confirmed" or a list of what was refused. One edit changes at most 200 cells. See [Tables](https://docs.usequotable.com/actions/tables.md#ask-quotable-agent).

> [accordion] Draft routines
>
> Ask for work that should repeat, for example "every Monday, summarise last week's visibility". Quotable Agent drafts a [routine](https://docs.usequotable.com/actions/routines.md) and shows it as a card with **Create**, **Edit first** and **Discard**. Nothing is saved until you confirm. It reads the brand's existing routines before drafting a change to one.

> [accordion] Act in your connected apps
>
> Use the tools of the MCP servers your brand connected, such as drafting a Contentful entry or updating a Framer collection. Tools that change something wait for your approval unless you allowed them. See [Approvals](#approvals) and [Connected apps](https://docs.usequotable.com/integrations/apps.md).

> [accordion] Connect an app
>
> Ask Quotable Agent to connect an account, such as Google Analytics, and it gives you a connect card. Accounts that are already connected are never reconnected from chat; Quotable Agent sends you to **Connectors** instead.

> [!NOTE]
>
> Quotable Agent cannot send email. Documents, briefs and articles are saved in
> the [Library](https://docs.usequotable.com/actions/library.md). For recurring work, use
> [routines](https://docs.usequotable.com/actions/routines.md).

## Questions before the work

When you ask Quotable Agent to make something without saying much about it, such as "write an article about our pricing" or "create a plan for Q1", it asks first. Up to four questions appear above the message box, one step at a time, each with likely answers drawn from your brand's data.

* Click an answer or press its number. Where more than one applies, the question says **Pick any that apply**: choose them, then click **Next** or **Done**.
* Type in the message box to answer in your own words.
* **Back** returns to the previous question. **Skip question** leaves the choice to Quotable Agent. **Skip all** tells it to go ahead with its best judgment.

After the last question, your answers are sent together and shown in the thread, and Quotable Agent starts the work. A request that already says who it is for and what format it takes goes straight to the work.

## Approvals

Reading data, searching, editing tables and saving documents happen without a prompt. **In chat, a connected app's tool set to Ask always waits for you.** Write and destructive tools start at Ask; you can change each one in [Tool permissions](https://docs.usequotable.com/integrations/apps.md#tool-permissions). A [routine](https://docs.usequotable.com/actions/routines.md) follows its own **What it may change** setting.

**1. Quotable Agent proposes the action**

An approval card shows the app, the tool and the complete content Quotable Agent wants to send.

**2. You decide**

Click **Approve** to run it once, or **Deny**. A denied step reads "Not approved."

* One call per approval, and Quotable Agent never chains calls from one approval.
* Each approval is single-use: it cannot be replayed.
* Quotable Agent cannot send email.

See [Connected apps](https://docs.usequotable.com/integrations/apps.md) for the servers you can connect and how to set them up.

## Attach files and context

Click **+** or type `/` in the composer.

| Attachment | Limits |
| - | - |
| Images (PNG, JPEG, WebP, GIF) | 10 MB each |
| PDF | 20 MB each |
| Text (TXT, MD, CSV, JSON) | 2 MB each |
| Documents from Quotable | Your tables (up to 200 rows) and indexed knowledge base documents |

Up to five attachments and 32 MB per message, with messages up to 200,000 characters. Quotable Agent reads up to 200,000 characters of attached document and text content per turn; an excerpt is marked when material exceeds that budget. Files are uploaded separately and kept privately for retries for 24 hours, then removed by the daily cleanup. The conversation keeps their names, not their bytes. If sending fails, the composer keeps your text and attachments so you can retry.

New turns settle near the top of the conversation. The chat follows a growing answer while you are caught up; scrolling upward keeps your reading position. Use **Latest**, above the composer, to return to recent messages. Image and PDF attachments have compact previews; PDFs use your browser's built-in viewer.

**Skills** are attached the same way, up to three per message. See [Skills](https://docs.usequotable.com/actions/skills.md).

## Model and effort

Quotable Agent picks a model for each task. The composer's **Effort** picker sets how much it reasons before answering: **Low** for simple tasks, **Medium** (default) for everyday work, **High** for complex tasks.

## Conversations

* Chats are private to the person who started them, per brand.
* The sidebar lists your chats from the last seven days. Chats are deleted automatically after seven days without activity.
* Quotable Agent uses the last 30 messages of a chat as context.

## Usage and limits

| What | Allowance |
| - | - |
| Chatting with Quotable Agent | Uses agent credits. Each answer shows the credits it used |
| Writes to connected apps | Use agent credits, as part of the answer that made them |
| Routine runs | Use agent credits; **Thorough** and **Deepest** use more |
| Rate limit | 60 Quotable Agent and report agent turns per hour per workspace |

Track credits in **Settings → Usage**. When a request does not fit the credits left, Quotable Agent refuses it before it starts. Credits reset on the 1st. When extra credits are on sale, the message and **Settings → Billing** offer them. See [Plans and limits](https://docs.usequotable.com/workspace/plans-and-limits.md).

Quotable Agent needs an active plan or trial, and edit access to the brand. Viewers cannot use Quotable Agent.

## Next steps

> [card] [Connected apps](https://docs.usequotable.com/integrations/apps.md)
>
> Connect MCP servers such as Contentful, Sanity and Framer.

> [card] [Routines](https://docs.usequotable.com/actions/routines.md)
>
> Turn a chat you repeat into a routine.

> [card] [Context Manager](https://docs.usequotable.com/actions/context-manager.md)
>
> Control the facts Quotable Agent uses.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.