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

# Remote MCP servers

> Connect a server the vendor runs, with no key to paste

Some services run their own MCP server. BoostGPT connects straight to it: you sign in, approve access, and
the tools arrive from the server itself. There is no key to paste and no app to register — BoostGPT
introduces itself to the server automatically.

> [!NOTE]
>
> These are an INPUT to what Mira learns, not a second place to work. They tell her what changed in the
> product, what visitors do, and what customers keep asking — so the Site and the Widget can say something
> truer.

## Reach the audience

| Server | What it gives Mira |
| - | - |
| **beehiiv** | Newsletter posts and subscribers. Which issues brought people back to the site. |
| **Customer.io** | Lifecycle messaging and segments. Use it to reach the people a page change is aimed at. |
| **Loops** | Audience, campaigns and transactional email. Use it to tell existing customers what changed. |
| **Resend** | Send email and read what happened to it. For the announcement that follows a page change. |

## Make images, video and voice for the site

| Server | What it gives Mira |
| - | - |
| **GitHub** | See what actually shipped. Read releases, merged pull requests and changelog files from PUBLIC repositories to find product changes the marketing site has not caught up with. Read-only, and prefer releases and merged PRs over raw commits — commit messages are mostly noise for this. |
| **Linear** | Read completed issues and cycles to see what shipped, in words a human wrote. A cleaner feed of product change than a commit stream. |
| **Airtable** | Bases and records. Wherever a team keeps the spreadsheet that runs part of the business. |
| **Sentry** | Errors and releases. Product health — it says what is broken, not what a page should say. |
| **Slack** | Public channels and their messages. Product context — what the team already said about a change; read-only, and never direct messages. |
| **Supabase** | Database, auth and storage for a product built on Supabase. Product context rather than marketing signal. |

## Research what people search for and ask

| Server | What it gives Mira |
| - | - |
| **Amplitude** | Product analytics — events, funnels and cohorts. Use it to check whether a page change moved the behaviour it was meant to move. |
| **Attio** | Who the leads are and what they have in common. Use it to find out which audience is actually converting before writing a page for a different one. |
| **DataForSEO** | What people search for, and who ranks for it. Use it to find the words customers use before they know your product exists. |
| **Front** | The shared inbox. Read recurring customer questions and confusion — the post-sale half of what the site should have explained. |
| **Help Scout** | Support conversations and saved replies. A saved reply written five times is a page the marketing site is missing. |
| **HubSpot** | Contacts, deals and the notes on them. Use it to see which objections come up before a deal closes, and what language customers use for the problem. |
| **Mixpanel** | Event analytics and funnels. Use it to see where visitors drop out between landing and signup. |
| **Paddle** | Subscriptions, prices and transactions. Use it to check whether a pricing-page change changed what people actually buy. |
| **Plain** | Customer conversations and threads. Read what customers ask after they sign up to find the gaps in what the site promised. |
| **PostHog** | What visitors actually do: pageviews, funnels, retention and session data. Use it to find which pages lose people and which paths convert. |
| **Stripe** | Revenue, subscriptions and churn. The end of the funnel: which pages and plans actually produce paying customers. |
| **Zendesk** | Support tickets and help-centre articles. Use recurring ticket subjects to find what the site failed to explain before signup. |

## Work on the site itself

| Server | What it gives Mira |
| - | - |
| **MetaManager** | Score how a page presents itself — title, description, Open Graph and Twitter cards — and get the exact fix. Use it on a page before and after a change to show the improvement. |
| **Contentful** | Entries, assets and content types. For marketing content managed in Contentful. |
| **Netlify** | Sites, deploys and forms. Use it to see what is live and to read form submissions from the site. |
| **Notion** | Pages and databases. Where positioning notes, launch plans and product context usually live. |
| **Sanity** | Structured content — documents, drafts and schemas. For content that lives in Sanity rather than in the site itself. |
| **Webflow** | Read and update a site already built in Webflow — pages, collections and CMS items. For a marketing site that lives outside BoostGPT. |
| **Wix** | Read and update a site built on Wix. For a marketing site that lives outside BoostGPT. |

## Connecting one that is not listed

You can point BoostGPT at any MCP server by its address — **Add your own** on the Connectors screen,
then **From an MCP server URL**. It is available on higher plans, because a server nobody has vetted
describes its own tools to your AI teammate that decides when to call them.

> [!WARNING]
>
> Only connect servers you trust. Mira reads a server's own description of its tools to decide when
> to use them, so a server you do not recognise could describe a tool misleadingly.


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