# Markup AI Content Guardian Agents

> Markup AI checks text for brand voice, terms, style, and risky claims. It works in Google Docs and Word, and an agent can send text through MCP or the API.

Markup AI checks written content against the brand voice, the terminology, and the style guide of a company. It calls its checks Content Guardian Agents. Each agent flags an issue, explains it, and suggests a fix for most issues.

The checks run in the tool where a person writes, for example Google Docs and Microsoft Word. An AI assistant or a program can also send text through MCP (Model Context Protocol) or the REST API and get the issues back.

| | |
| --- | --- |
| Group | Automatic checks |
| Where the work must be | Markup AI checks text in the tool where a person writes, for example Google Docs, Microsoft Word, and Figma. It also checks text that an AI assistant or a program sends through MCP (Model Context Protocol) or the API. In Google Docs, Markup AI works through a Chrome extension that adds a side panel. The extension needs a Chromium browser, for example Chrome or Edge. In Figma, Markup AI runs as a plugin. The user selects the text layers, and the plugin checks only that selection. |
| Can an agent run it? | Yes. Markup AI hosts an MCP (Model Context Protocol) server at https://api.markup.ai/mcp/. An agent sends text to the markupai_review tool or the markupai_run_agents tool and gets the issues back. Yes. The REST API runs an agent with POST /agents/{agent_id}/run. The caller passes the content as text and gets back the flagged issues with suggested fixes. An MCP (Model Context Protocol) client signs in with a Markup AI account through OAuth or sends an API key. A Markup AI account is necessary. |
| File types | The MCP (Model Context Protocol) tools take the content as text. Each call accepts up to 100,000 characters. The Style Agent API also accepts a file: PDF, Word (.docx), HTML, XML, Markdown, or plain text. The maximum file size is 7 MB. |
| What the check gives back | The Style agent gives the location of each issue, an explanation, and a suggested correction. It also gives quality scores for each goal. Each issue has a risk level of High, Medium, or Low. All agents use the same scale. The MCP (Model Context Protocol) tools also report which agents ran and which voice profile, style guide, and persona they used. Not every issue has a suggested fix. In Figma, an issue that is for awareness only shows the flagged text and the reason. |
| Who decides | The check advises. For each issue in Google Docs, the writer applies the fix, adds it as a suggestion or a comment, or dismisses it. The GitHub Action can gate a merge. It posts a commit status of success, failure, or error from the overall risk, and branch protection can use that status. |
| Where the rules come from | An organization sets its brand voice, personas, terminology, and style guide in the Markup AI console. Each agent reads those settings when it runs. On the Starter and Business plans, an Administrator configures style guides, terminology, and the Brand Profile. A Writer can see them but cannot change them. For the Style agent, the user picks the Default style guide from Markup AI or a custom style guide of the organization. For the Brand Terms agent, a term set holds prohibited terms, preferred replacements, and guidance on when each rule applies. Markup AI offers a free Brand Profile that it builds from the public content of a company. |
| Brands in one account | The Business plan includes multiple brand profiles. Through MCP (Model Context Protocol), a call can name the brand voice profile, the style guide, and the persona to use. |
| Plan with the check | The Starter plan includes the review for brand, quality, and citability. The Business and Enterprise plans include everything in Starter. The agents that an organization can run depend on its plan and its feature access. The Style Agent is not in every plan. If the Style guides tab is missing from the groups, the plan does not include the Style Agent. The pricing page lists Figma, VS Code, Contentful, and AEM Guides only under the Enterprise plan. Starter and Business list Google Docs, Microsoft Word, the API, ChatGPT, Claude, and MCP (Model Context Protocol). |
| Free plan | The pricing page offers a free 30-day trial. The docs page Plans & Trials says that every organization starts with a 7-day free trial. The trial needs no credit card and gives the full product. If the trial ends, access pauses until the organization chooses a plan. Markup AI keeps the style guides, terminology, and settings. |

## What does Markup AI check?

| Check | Result | What the vendor says |
| --- | --- | --- |
| Logo | No. Text only | No. Markup AI checks text. Its guide for Google Docs says that it does not analyze images, charts, and other non-text elements. “images, charts, and other non-text elements aren’t analyzed” |
| Color | No. Text only | No. The Figma plugin extracts the text from the selected layers and checks that text. It ignores images, shapes, and other non-text content. “Markup AI focuses on the text in your layers; images, shapes, and other non-text content are ignored.” |
| Type | No. Text only | No. The Figma plugin extracts the text from the selection and analyzes that text. The guide lists non-text elements as not analyzed. “The plugin extracts the text from the selection, analyzes it, and posts the result back to the sidebar.” |
| Copy and terms | Yes | Yes. The Brand Voice agent finds passages where the tone, personality, or phrasing drifted from the approved brand voice. It gives a suggested rewrite for each passage. Yes. The Brand Terms agent checks content against the terminology of the organization. It covers the words to avoid and the exact form of each name. Yes. The Style agent checks content against the style guide of the organization. Its goals are Clarity, Consistency, Inclusive language, Scannability, Spelling and grammar, Terminology, and Tone. “Passages whose tone, personality, or phrasing has drifted from your approved brand voice” |
| Claims and legal lines | Flags risky claims | Yes. The Claims agent flags bold or unsubstantiated claims that can carry compliance risk. It labels each claim with its risk type. The Claims agent looks for strong outcome claims without evidence and for promissory language. A different agent, Source Authority, covers missing attribution. “This agent flags bold or unsubstantiated claims that may carry compliance risk.” |

## The fine print

- Markup AI checks text only. In Google Docs, it does not analyze images, charts, and other non-text elements. The vendor's words: “Checks cover text content; images, charts, and other non-text elements aren’t analyzed.” ([source](https://docs.markup.ai/apps/google-docs))
- In Figma, only the Style Agent runs. The guide says that the agents under Compliance, Content Integrity, and AI Visibility are previews. The vendor's words: “the agents under Compliance, Content Integrity, and AI Visibility are previews” ([source](https://docs.markup.ai/apps/figma))
- In Figma, a check cannot cover a whole page or file. The user must select the text layers. The vendor's words: “Can I check my whole Figma page or file at once?” ([source](https://docs.markup.ai/apps/figma))
- In Word, Markup AI does not analyze text boxes and shapes. It does not support SmartArt, WordArt, charts, and OLE objects. The vendor's words: “their content isn’t extracted, so it’s neither analyzed nor modified” ([source](https://docs.markup.ai/apps/m365-word))
- Many apps run only the Style agent. The App Directory marks VS Code, Contentful, AEM Guides, Oxygen XML, XMetaL Author, MadCap Flare, Heretto CCMS, and Figma as Style agent. It marks Google Docs, Google Slides, Word, and PowerPoint as All agents. ([source](https://docs.markup.ai/apps))
- The Brand Terms agent checks a very long document only up to the first 1,000 matches. The result states this. The vendor's words: “Very long documents are checked up to the first 1,000 matches, and the result says so.” ([source](https://docs.markup.ai/agents/terminology))
- By default, the Brand Terms agent does not flag quoted text, titles and names of works, link text, and URLs. The vendor's words: “Any of these can be switched on for a term through its guidance.” ([source](https://docs.markup.ai/agents/terminology))
- MCP (Model Context Protocol) runs are synchronous, and a very long document can time out. MCP has no async mode and no polling. The vendor's words: “Runs are synchronous and bounded by a server-side deadline, so very long documents can time out.” ([source](https://docs.markup.ai/mcp))
- Each plan has a yearly allowance of checked words. Markup AI says that the allowance is a soft limit and that it does not interrupt a check. The vendor's words: “Your annual word allowance is a soft limit.” ([source](https://docs.markup.ai/admin/billing/plans))

## Which plans does Markup AI have?

- Starter is for marketing teams of 1 to 3 people. It includes centralized administration, the review for brand, quality, and citability, suggested improvements, and community support. The pricing page shows no price and has a Book a Call button.
- Business is for teams of 4 or more people. It adds Content Audit for batch checking, multiple brand profiles, and priority support by ticket. The pricing page shows no price and has a Book a Call button.
- Enterprise adds custom word volume, custom review criteria, the complete list of integrations, single sign-on, and a dedicated Customer Success Manager. The pricing page shows no price and has a Book a Call button.
- The docs say that Starter and Business are self-serve plans with a flat price and unlimited users. Markup AI bills them monthly or annually.
- Markup AI does not sell the Enterprise plan online. The Markup AI account team manages the contract and the billing.

Prices are from [the vendor's pricing page](https://markup.ai/pricing/).

## What else do Markup AI's pages say?

- The App Directory lists apps for Google Docs, Google Slides, Word, PowerPoint, VS Code, Contentful, AEM Guides, Oxygen XML, and XMetaL Author. It also lists MadCap Flare, MadCap Flare Online, MadCap IXIA CCMS, Heretto CCMS, Confluence Cloud, and Figma. ([source](https://docs.markup.ai/apps))
- The docs have MCP (Model Context Protocol) setup guides for Claude Desktop, ChatGPT, Claude Code, Cursor, Codex, Kiro, VS Code, Windsurf, Zed, and LibreChat. ([source](https://docs.markup.ai/mcp))
- The Recipe Gallery has guides for Zapier, n8n, MCP (Model Context Protocol), and a GitHub Action. The GitHub Action analyzes changed files on pull requests and pushes. ([source](https://docs.markup.ai/recipes))
- In Figma, an audit trail popover shows the totals for each severity and the counts of applied and dismissed suggestions for the run. ([source](https://docs.markup.ai/apps/figma))
- Content Audit checks content that is already published. A user uploads files, crawls a website, or connects Google Drive, and gets a score and findings for each document. ([source](https://markup.ai/product/))

## Who does it suit?

It suits a team whose work is mostly text: articles, product pages, help content, and email. Writers see the issues in their document, and an agent that writes copy can check its own draft before a person reads it. The GitHub Action can also stop a merge when the risk is high.

## When is another product a better fit?

- If your work is visual, look at [Rainbrand](https://brandcompliancesoftware.com/directory/rainbrand), [Bynder](https://brandcompliancesoftware.com/directory/bynder), or [Lytho](https://brandcompliancesoftware.com/directory/lytho). Markup AI checks text only, and its guides say that it does not analyze images.
- If people must sign off each piece, add a proofing tool such as [Filestage](https://brandcompliancesoftware.com/directory/filestage) or [Ziflow](https://brandcompliancesoftware.com/directory/ziflow). Markup AI advises the writer.
- If you want to see a price first, note that the pricing page shows no price. Each plan has a Book a Call button.

## Compare

- [Markup AI vs Writer](https://brandcompliancesoftware.com/compare/markup-ai-vs-writer)

## Where are the vendor's own pages?

- [Markup AI product page](https://markup.ai)
- [Documents](https://docs.markup.ai)
- [Pricing page](https://markup.ai/pricing/)

## Sources

- <https://markup.ai/> (checked 2026-11-16)
- <https://markup.ai/product/> (checked 2026-11-16)
- <https://docs.markup.ai/agents> (checked 2026-11-16)
- <https://docs.markup.ai/apps/google-docs> (checked 2026-11-16)
- <https://docs.markup.ai/apps/figma> (checked 2026-11-16)
- <https://docs.markup.ai/agents/brand-voice> (checked 2026-11-16)
- <https://docs.markup.ai/agents/terminology> (checked 2026-11-16)
- <https://docs.markup.ai/agents/style-agent> (checked 2026-11-16)
- <https://docs.markup.ai/agents/generic-claims> (checked 2026-11-16)
- <https://docs.markup.ai/mcp> (checked 2026-11-16)
- <https://docs.markup.ai/developer-reference/api-reference/style-agent/run-style-agent-file> (checked 2026-11-16)
- <https://docs.markup.ai/apps> (checked 2026-11-16)
- <https://docs.markup.ai/quickstart> (checked 2026-11-16)
- <https://docs.markup.ai/github-action> (checked 2026-11-16)
- <https://docs.markup.ai/admin/roles-and-permissions> (checked 2026-11-16)
- <https://markup.ai/brand-profile> (checked 2026-11-16)
- <https://markup.ai/pricing/> (checked 2026-11-16)
- <https://docs.markup.ai/developer/agents> (checked 2026-11-16)
- <https://docs.markup.ai/admin/style-agent> (checked 2026-11-16)
- <https://docs.markup.ai/admin/billing/plans> (checked 2026-11-16)
- <https://docs.markup.ai/admin/billing/manage> (checked 2026-11-16)
- <https://docs.markup.ai/apps/m365-word> (checked 2026-11-16)
- <https://docs.markup.ai/recipes> (checked 2026-11-16)

---

Source: https://brandcompliancesoftware.com/directory/markup-ai
Last updated: 2026-11-16
Published by Fine Print. Fine Print is run by Rainbrand, which makes one of the products listed. Its entry follows the same rules as the others.
