Reading as an AI? Append .md to this URL for markdown — /llms.txt
Back to Docs

Integrations

External apps your AI agents can plug into to take real actions — answer order questions, send emails, update stores, look up records. Connect once, then decide which agents can use it and what they are allowed to do.

Available integrations

Open any of these for the per-app guide. The connect flow and the agent-permission controls work the same way across all of them.

Assigning agents and setting permissions

Every integration starts isolated — connecting it does not automatically give any AI agent access. You explicitly assign agents to each connection and pick what they can do. The workflow is the same across every app.

1Open the app's management page

Sidebar → Connect → Already connected. Each app you signed into appears in the list. Click one to open its management page (right pane on desktop, separate page on mobile).

Permission step 1: Open the app's management page

2Click + Add AI Agent and pick the agents

A popup lists every AI agent you own, with each one's profile picture. Tick one or several — you can add multiple at once. Each agent is added with access Off by default, so nothing happens until the next step.

Permission step 2: Click + Add AI Agent and pick the agents

3Manage each agent from the list

Every agent you added appears in Agents using this app with its current preset. Two controls per row: a trash icon removes the agent from this app (the agent itself is untouched), and an Edit button opens the per-action permissions modal.

Permission step 3: Manage each agent from the list

4Set the access level and override individual actions

Edit opens the permissions modal. Pick the access level at the top: Off, Read only (search/read but never modify), or Read and write (full access). The preset sets the default state for each action; flip individual ones on or off to deviate from it — e.g. Read only + Create an order ON gives an order-taking agent that cannot delete anything. Save when done; changes apply on the next message the agent processes.

Permission step 4: Set the access level and override individual actions

Access levels at a glance

  • Off — the agent stays linked but cannot use the app. Use it to temporarily revoke access without losing the configuration.
  • Read only — the agent can search, list, and read; never modify or send. Good for triage agents.
  • Read and write — the agent can do everything the app exposes. Use for fully autonomous workers.

Permissions are enforced server-side on every action — the UI is not the gate. Even if a user prompt asks the agent to do something the preset forbids, the action is refused at the tool layer.

Multiple accounts of the same app

You can connect more than one account of the same app (e.g. a work Gmail and a personal Gmail, or a retail Shopify store and a wholesale store). Each connection has its own row in Already connected, and you assign agents to each one separately.

When an agent has access to two or more accounts of the same app, it picks the right one from context — or asks the user if the request is ambiguous. The optional Note field on each connection ("Sales inbox", "Wholesale store") gives the agent a human label to anchor on.

Common surprises

  • "The agent says it cannot use the app" — its preset is Off, or it is set to Read only and the user is asking for a write action. Open Edit and adjust.
  • Removing vs disconnecting — removing an agent from the list revokes only that agent's access. Disconnecting the app at the top wipes every agent's access in one shot.
  • Re-adding an agent — when you re-add an agent you previously removed, it starts at Off. The old per-action overrides are not remembered.

Privacy

Kronos does NOT store the content of your integrations — orders, products, customers, email bodies. Each time an agent uses an action it fetches just-in-time, uses what it needs, and discards the result. The audit log records that the action happened (which agent, which action, when) but not the underlying data.

When you disconnect an app or cancel your Kronos subscription, the connection and every per-agent permission row on it are deleted immediately. See the Privacy and Data deletion pages for the legal and processor details.

For AI agents and LLMs: every docs page has a clean markdown twin — append .md to its URL. Curated index: /llms.txt. Open semantic search, no auth: POST https://api.kronos.com.mx/v1/docs/search.
Integrations — Docs — Kronos AI | Kronos AI