Claude
- Open Settings, then Connectors, and choose Add custom connector.
- Name it Koltrix and paste https://mcp.koltrix.com/mcp as the URL.
- Click Connect, sign in to Koltrix, and approve.
- In a chat, turn Koltrix on from the tools menu.
AI assistants (MCP)
Connect Koltrix to Claude, ChatGPT, Cursor or any app that speaks MCP, the open standard for giving AI assistants tools. You sign in and approve it once; it sees only the mailboxes you can, and an admin can switch it off for the whole workspace.
The server URL. This is all you paste, there is no API key.
https://mcp.koltrix.com/mcpWhat needs a reply from me today?
Used get_inbox_summary in Koltrix
Two conversations are waiting on you:
Draft a reply to Dana: yes, we ship Friday.
Used create_draft in Koltrix
Draft, not sent
Hi Dana, yes, we ship on Friday. I will send tracking as soon as it leaves.
The promise
Claude and ChatGPT can read, search, organise and draft. Sending is off by default: an admin can turn it on, and the assistant must show you the message and get your yes before it sends.
Koltrix's own AI, the agents that sort, summarise and suggest replies inside the app, is separate and unchanged: it never sends without a person clicking Send.
What needs a reply from me today?
Used get_inbox_summary in Koltrix
Two conversations are waiting on you:
Draft a reply to Dana: yes, we ship Friday.
Used create_draft in Koltrix
Draft, not sent
Hi Dana, yes, we ship on Friday. I will send tracking as soon as it leaves.
Connect
Add Koltrix as a connector in your assistant. There is no API key to create or copy.
Your assistant sends you to Koltrix. Choose the workspace, read the permissions in plain words, click Approve.
Turn Koltrix on in a chat and talk to your inbox: search, summarise, tidy up, draft.
Menu names change between plans and releases; the setup guide in the docs carries the current ones.
{
"mcpServers": {
"koltrix": {
"url": "https://mcp.koltrix.com/mcp"
}
}
}claude mcp add --transport http koltrix https://mcp.koltrix.com/mcpAny other app that supports remote MCP servers over Streamable HTTP with OAuth sign-in connects with the same URL. Full setup guide.
Permissions
The approval page at app.koltrix.com lists each permission in plain words. The connection belongs to you, in one workspace.
mail.read
Search and read your mail, and list your mailboxes and labels.
mail.organize
Add and remove labels, archive, mark read or unread, and star. All of it can be undone in Koltrix.
mail.draft
Save drafts to your Drafts folder for you to review. On its own it cannot send them.
mail.send
Off unless an admin allows it, then an unticked box you choose to tick when you connect. Needs your yes on every message.
Apps ask for the first three by default. The send permission is never part of that default: it appears as its own unchecked box, and only when an owner or admin has turned assistant sending on.
Under your control
Anyone can send you an email, and an email can contain text written to fool an AI assistant. So Koltrix does not rely on the assistant behaving: the limits are enforced on our side, on every call.
An assistant sees only the mailboxes you can open in Koltrix, checked on every call.
See every connected app in your settings and cut one off. It stops working immediately.
Owners and admins can turn assistants off for the whole workspace, and sending is a second switch that starts off.
There is no tool that deletes mail or forwards it. Labels, archive, read state, stars and drafts can all be undone.
At most 10 recipients a message, 20 sends a day per connection and 50 a day per workspace through assistants. No Bcc, no attachments.
Connections, revocations and tool calls are recorded in the workspace audit log with the tool name, never your mail.
Sending, when allowed, works on a draft that already exists, so the exact message is in your Drafts folder. The server refuses unless the assistant names the real recipients and subject, and it marks the call as irreversible so apps such as Claude and ChatGPT ask for their own approval too. Koltrix cannot see your conversation, so read the message before you say yes. How sending works, in full.
Interactive cards
Some assistant apps can show a tool's result as a small interactive card, using the MCP Apps extension. An app that does not support cards gets the same tools with text results. Koltrix offers cards only to an app that says, when it connects, that it supports them.
list_mailboxes
Choose which mailbox the assistant drafts from.
search_mail, list_threads, get_inbox_summary
Sender, subject, snippet, time, unread dot and labels, with a link to open it in Koltrix.
get_thread
Read the messages as plain text, attachments listed by name and size.
create_draft
See From, To, Cc, Subject and the whole message, then Send, Edit in Koltrix or Discard.
Anthropic documents card support for Claude on the web, in Claude Desktop and on mobile, and OpenAI documents it for ChatGPT. Claude Code shows text. Whether a given app draws a card is decided by the app. The cards load nothing from the internet and show anything that came from an email as plain text.
The Model Context Protocol is an open standard for giving AI assistants tools. A service such as Koltrix publishes a list of things an assistant may do, and any app that speaks MCP, such as Claude, ChatGPT or Cursor, can use that list once you have signed in and approved it.
Only if you choose to allow it. Out of the box assistants read, search, organise and draft. A workspace owner or admin can turn sending on; then you opt in when you connect, and the assistant has to show you the recipients, subject and whole message and get your yes before it sends. Limits apply: 10 recipients, 20 sends a day per connection and 50 a day per workspace, no Bcc and no attachments. No assistant can forward or delete mail. The full rules are at https://docs.koltrix.com/mcp.
No. You paste one server URL and sign in to Koltrix when the app asks. The connection belongs to you, in one workspace, and uses standard OAuth sign-in. To use a second workspace, connect again and choose it.
The mailboxes you can open in the Koltrix app and nothing else, checked on every call. A teammate's mailbox you have no access to stays invisible to your assistant too.
The parts of your mail the assistant reads are sent to that assistant's provider (Anthropic for Claude, OpenAI for ChatGPT) and handled under their terms. Koltrix sends only what a tool call asks for and does not receive your conversations with the assistant, only the tool calls it makes.
Revoke a single app in your Koltrix settings and it stops working immediately. Owners and admins can also turn assistants off for the whole workspace, which cuts every connection at once.
No. Koltrix's own AI sorts mail, writes summaries and suggests replies inside the app, and it never sends without a person clicking Send. MCP is for assistants you already use elsewhere, working through your Koltrix account.
Read more
Add your domain, connect your assistant and ask what needs a reply. Every plan starts with a free trial, no card.