Connected agents and webhooksAvailable

Connect your agents to Sixto

A connected agent works through the Sixto API with a key and, when it acts for a person, it never exceeds that person's permissions. Sensitive actions, such as messaging guests, invoicing, touching a lock, or changing company email, wait for a person to confirm them in Sixto.

Available. Connect your own AI agent or automation to Sixto with a scoped key and signed webhooks, under Settings > Integrations > Agents.

Sixto Connected agents and webhooks
CONNECTION WORKFLOW

A connected agent, clear boundaries

Scoped keys and signed webhooks

SixtoEvent with permitted data
Connected agentYour chosen permissions
ActionRecorded in the activity log
Illustrative workflow. Example, not a live account view.
[ WHAT IT DOES ]

What it does

  • Your agent, your name

    Requests on your behalf use your permissions. Standalone API keys retain their selected scopes, and supported changes still follow confirmation settings.

  • Signed events

    Events are sent to your webhook URL, signed so you can verify them, and retried until delivered. Duplicates carry the same event ID.

  • Acting on behalf of a person

    When you ask a connected agent, it inherits your current permissions and property access, including company email. Each call checks your access again. Existing requests keep their original limits; new permissions apply to new requests.

  • Read context when it is needed

    During an active request, your agent can look up product instructions and permitted records, including the company email inbox, invoices and accounting summaries. Every read follows the request’s access limits and stops when the request closes.

  • Routing you can see

    When someone asks Sixto in chat, Sixto shows whether it handled the request with platform tools, passed it to your agent or reported it to the Sixto team.

  • No loops

    Your agent's own changes never come back to it as events.

[ MCP ]

Connect AI apps with MCP

Sixto has an MCP server, so an AI app can read and work with your Sixto data with the same permissions as the person who connects it. It stays off until you turn it on.

Server address
https://sixto.app/mcp
Docs MCP
https://sixto.app/mcp/docsPublic and read only. It searches Sixto's documentation and never sees your organization's data.
  • Turn it on first

    An Owner or Manager opens Settings > Integrations and turns on Allow AI apps (MCP). It is off by default. If the card says AI apps are switched off in the platform console, ask the Sixto team to switch them on for your organization.

  • Who can connect

    Only the Owner and Manager roles. Staff, accountants, cleaners and read-only roles cannot connect an app, and the app never gets more than the permissions of the person who connected it.

  • Sign in with Sixto

    Add the server address in your AI app. It opens Sixto: sign in with your Sixto account, choose the organization and what the app may do, and press Allow.

  • Confirm changes in Sixto

    Sending a guest message, an e-invoice or a pay link, guest door codes and other sensitive changes come back as a confirmation card. A person confirms it in Sixto; the AI app cannot. Routine team changes, such as creating a task, run within the person's permissions.

  • Revoke at any time

    Settings > Integrations > Connected apps lists every AI app you allowed. Revoke cuts the app off on its next request.

Works with AI apps that can add a remote MCP server with sign-in (Streamable HTTP and OAuth), such as Claude and ChatGPT.

[ HOW IT WORKS ]

How it works

Sixto Connected agents and webhooks
CONNECTION WORKFLOW

The connection at a glance

Scoped keys and signed webhooks

SixtoEvent with permitted data
Connected agentYour chosen permissions
ActionRecorded in the activity log
Illustrative workflow. Example, not a live account view.
  1. 01

    An Owner creates a key for the agent and picks its permissions in the API keys area.

  2. 02

    Under Settings > Integrations > Agents, enter your agent's webhook URL and a shared secret. The secret is stored encrypted.

  3. 03

    Your agent verifies each event's signature, answers quickly and does the work afterwards through the API.

  4. 04

    Trim or remove the agent's permissions any time. The next call is refused.

[ DATA ]

What syncs

DataDirectionDetails
Events (chat, feedback, tasks, pricing requests)From SixtoSixto to your agent's webhook, signed, with free text excerpted and masked.
ActionsInto SixtoYour agent to Sixto through the API, within its permissions.
[ PRIVACY ]

Permissions and privacy

  • Events are only sent to your own organization's agent, never to another organization's.

  • Free text is bounded and secrets are redacted. Guest contacts are available only when the person asking may view them.

  • An agent never gets more permission than the person it acts for, and it can never unlock a door, approve messages or manage integrations and team members.

  • Nothing is sent until you set a webhook URL and secret.

[ FAQ ]

Common questions

Do I need an agent?

No. Sixto's own assistant works without one. Connecting an agent is optional.

Which agents work?

Any agent or automation that can receive a signed webhook and call the REST API. ChatGPT and Claude connect to the MCP server at https://holasixto.com/mcp: an Owner or Manager turns on Allow AI apps (MCP), signs in, chooses the organization and scopes, and presses Allow. Sensitive actions wait for Confirm in Sixto. Revoke the app under Settings > Integrations.

What if my agent is down?

Events are retried, and the agent can read what it missed through the API.

Connect what you use. Skip what you do not.