Module · switch on per organisation
Connect your AI assistant, safely
Enclessa lets the AI assistant somebody already uses — Claude, an IDE, any client that speaks the Model Context Protocol — connect to their organisation by URL. The person signs in in their browser and the assistant then acts as them: it sees exactly what they may see and nothing more, each tool is set to off, ask first, or act without asking, and every change it makes is written to the audit log under that person’s name. Connected apps are listed in the person’s settings and can be ended there at any time.
Updated
- Availability
- Module, off by default
- Protocol
- Model Context Protocol, OAuth2 with PKCE
- Identity
- The assistant acts as the signed-in person
- Per tool
- Off, ask first, or act without asking
- Revoking
- Each connected app listed and ended in your settings
What can a connected assistant reach?
Exactly what the person who connected it can reach, and less where the organisation says so. Tools for modules the organisation has not installed, or for permissions the person does not hold, are absent rather than refusing. Whether an assistant may read the text of messages, mail, documents, knowledge-base pages or onboarding answers is five separate switches, all off by default: off, it can find where something is without reading what it says.
- Signs in as the person, through their browser, with OAuth2 and PKCE
- Sees only what that person may see, across the installed modules
- Reading message, mail, document, page and onboarding text is off until switched on
- Every change written to the audit log under the person’s name
Who decides what it may change?
The organisation, per tool, and the person within that. Each tool is off, ask first, or act without asking; every tool that changes something starts at ask first, so the assistant drafts the change and a person approves it before anything happens. The connection itself belongs to the person: their settings list every app they connected, and ending one cuts it off at the next call.
- A setting per tool: off, ask first, or act without asking
- Every tool that changes something starts at ask first
- Connected apps listed in account settings, each one revocable
- The same rules as the Companion, the assistant built into Enclessa
What it does not do
Stated here rather than discovered after signing up.
- AI agents are a module an administrator switches on for the organisation. It is included in the Business plan and above.
- An assistant cannot read end-to-end encrypted rooms. The server holds no key for them, so no tool can reach their content.
- The assistant itself runs at its own provider. What it is shown leaves Enclessa under that provider’s terms, which is why reading message and document text is off until an organisation switches it on.
Related
Where this shows up
Questions
AI agents: common questions
- Yes, with the AI agents module. Claude, or any assistant that supports the Model Context Protocol, connects by URL; you sign in in your browser, and the assistant then acts as you, seeing only what you may see.
- Only where the organisation sets a tool to act without asking. Each tool is off, ask first, or act, and every tool that changes something starts at ask first, so a person approves the change before it happens.
- In your account settings. Every app you connected is listed there, and ending one revokes its access at its next call.
Your team. Your keys. Your continent.
Create an organisation in a minute. Free for up to 10 active members, forever.