> For the complete documentation index, see [llms.txt](https://missiveapp.com/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://missiveapp.com/docs/ai/mcp/server/permissions.md). # Permissions and scopes What each MCP permission grants, which actions it unlocks, and how to give an AI app only the access it needs. When an AI app connects to Missive, it asks for a set of permissions. You approve them one by one on the authorization screen, and you can uncheck any of them. Permissions are enforced, not advisory. Missive hides the matching capabilities from the app completely, so an app without **Read calendars** doesn't get a calendar tool it's trusted not to use. It never sees one. ## The permissions | Permission | What it grants | | ------------------------------- | -------------------------------------------------------------------------------- | | **Read conversations** | View your conversations and the messages, comments and attachments they contain. | | **Manage conversations** | Organize and make changes to your conversations on your behalf. | | **Read contacts** | View your contacts and their details. | | **Read organizations settings** | View your organizations, teams, shared labels and other organization settings. | | **Create drafts** | Compose and create message drafts on your behalf. | | **Send messages** | Deliver message drafts to their recipients on your behalf. | | **Read calendars** | View your calendars and their events. | | **Manage calendars** | Create and edit events on your calendars. | Drafting and sending are deliberately separate. An app can write replies for you to review without any ability to put them in front of a recipient. If you want a human check before anything goes out, grant **Create drafts** and withhold **Send messages**. ## What each permission unlocks Most of what an AI app does through Missive is read-only. The write actions are listed separately below so you can see exactly what you're handing over. ### Read-only | Permission | Actions | | ------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------- | | **Read conversations** | Search conversations across all connected accounts, fetch conversations by label, team, or account, and load older messages in a long thread. | | **Read contacts** | Look up contacts by email, phone, or keyword. | | **Read organizations settings** | Search mailboxes, shared labels, and teams, and search your canned responses. | | **Read calendars** | List the calendars you can access and read events and busy times to check availability. | ### Write | Permission | Actions | What to watch | | ------------------------ | -------------------------------------------------------------------------------- | ---------------------------------------------------------------- | | **Manage conversations** | Add and remove shared labels and mailboxes on a conversation. | Changes your team's shared organization, not just your own view. | | **Create drafts** | Create and update drafts, whether a reply, a new message, or a new conversation. | Drafts stay unsent until something delivers them. | | **Send messages** | Deliver a draft to its recipients. | Sends real email as you. | | **Manage calendars** | Create, update, and delete events, including video conferencing links. | Deleting an event notifies attendees. | {% hint style="warning" %} Write actions taken through MCP happen immediately. Missive's own in-app confirmation step doesn't apply here, so the only thing standing between a request and a sent email is whatever approval prompt your AI app shows you. Check that your app is set to ask before running tools, especially if you granted **Send messages**. {% endhint %} ## What an app can reach An app acts as you. It sees what you see in Missive and nothing more: your conversations, your accounts, your calendars, and the shared data of the teams you belong to. Missive applies the same access checks it applies in the app itself. One consequence worth understanding: your access spans **every organization you belong to**, so a connected app does too. That's why MCP has to be enabled in all of them before you can connect at all. See [Admin controls](/docs/ai/mcp/server/admin.md#multiple-organizations). ## Choosing what to grant A few sensible starting points: | You want to | Grant | | -------------------------------------------- | ------------------------------------------------------------------------------ | | Ask questions about your inbox and calendar | Read conversations, Read contacts, Read organizations settings, Read calendars | | Also have it write replies for you to review | Add Create drafts | | Also let it triage and label | Add Manage conversations | | Let it handle a thread end to end | Add Send messages | Start narrow. Adding a permission later means [revoking the app](/docs/ai/mcp/server/connected-apps.md#revoking-an-app) and reconnecting, which takes under a minute. ## Related * [Connecting an AI app](/docs/ai/mcp/server/connect.md) * [Managing connected apps](/docs/ai/mcp/server/connected-apps.md) * [Admin controls](/docs/ai/mcp/server/admin.md) --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://missiveapp.com/docs/ai/mcp/server/permissions.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.