Cloud Intelligence™ pulls from your team's entire tech stack when running AI-assisted FinOps investigations.

Connect Linear, GitHub, Notion, Atlassian, or any MCP server, and the AI assistant works in them from the same conversation as your cost data.

An anomaly shows up in Cloud Intelligence™. The AI assistant tells you which service spiked, in which account, and roughly what it will cost by month end. Acting on it has meant leaving. Open Linear or Jira, check whether someone already filed it, copy the numbers over, tag the owning team, come back. The assistant could do the analysis. The follow-through was yours, by hand, in another tab.

You can now hand it the follow-through too. Cloud Intelligence™ connects to any remote Model Context Protocol (MCP) server, and the assistant uses that server's tools alongside its own.

What you get

A Cloud Intelligence™ administrator connects a server once from Integrations. Pick Linear, GitHub, Notion, or Atlassian from the presets, or paste the URL of any MCP server your own team runs. The platform contacts the server, works out how it authenticates, and reads its tool list. From then on the assistant picks connected tools the same way it picks its built-in ones. Ask "Is there already a Linear issue about yesterday's EKS anomaly? If not, open one for the platform team." It searches first, then drafts the issue with the cost context already filled in.

Each person authorizes their own account with the third-party service. The assistant acts with that user's permissions, and every action is attributed to them. If it needs a server you haven't authorized yet, it shows an Authorize card in the chat. Complete the sign-in, select Retry, and it carries on. In Slack the same moment becomes an Authorize button that opens the console.

You can also steer. Select Add > Connected tools in the composer and pick one or more servers. They stay attached to the conversation as chips. The assistant then treats your request as being about those systems and reaches for their tools first.

Setup is deliberately boring. There's nothing to install and no shared service account to create. The MCP servers guide covers the whole thing in a few screens.

Guardrails

Administrators decide which tools the assistant may touch, per tool or per group. Tools marked read-only run without asking. Write tools are on by default and ask before running. The assistant shows the exact request and waits for Run or Cancel, in the Cloud Intelligence™ UI, in Slack and the DoiT MCP server. Destructive tools stay off until an administrator turns them on. Tools that appear after a server update are marked New and keep their group's default until someone reviews them. Anything a connected server returns is treated as data, so the assistant never follows instructions found inside a tool result.

Get started

1. Connect a server from Integrations > MCP servers. You need the "Manage settings" permission.
2. Choose which tools the assistant may use in the server's details.
3. Authorize your account, from the server row or the first time the assistant asks.
4. Ask the assistant as you normally would, or pick servers from Add > Connected tools to point it at them.

MCP servers are available now to all Cloud Intelligence™ customers. Start with the tracker your team already files cost work in. The feedback we want most is from the first time the assistant opens an issue for you. Tell us what it got right, and which server you wish worked and doesn't yet.