This week in Maia: Mission Control and Context Engine

Welcome back to the Maia New Features Blog! This week, we're thrilled to announce the public preview launch of Mission Control and Context Engine—two major new capabilities that transform how you manage AI agent tasks and ground Maia AI Agents in your business data. For a full list of changes this week, check out our changelog updates.

🗺️ Context Engine: give Maia AI Agents a living map of your data

Context Engine is now in public preview! This powerful new feature enables you to builds knowledge graphs—a living map of the business logic, entities, and relationships in your data landscape—so Maia AI Agents can generate accurate, production-ready pipelines that speak the language of your business.

Maia AI Agents are only as good as their understanding of your data and your business. Out of the box, they know the shape of a warehouse, but not what the tables mean, how concepts relate, or which data is trustworthy. Context Engine closes that gap by crawling your warehouse, inferring the business entities behind your schemas, and storing them as a semantic graph that Maia AI Agents can draw on. The result is less hand-holding, fewer wrong guesses, and pipelines that match how your organization actually thinks about its data.

What's new:

• Domain-focused knowledge graphs: Create a graph per domain (Finance, Sales, Marketing...). Keeping each graph focused helps Maia AI Agents learn the language and logic of that data far more effectively.
• Automated crawlers: Attach a crawler to a warehouse connection and it harvests physical assets and infers the business concepts behind them, keeping the graph accurate and in sync. Works across Snowflake, Databricks, and Amazon Redshift.
• Pipeline execution crawls: Bring in runtime lineage and execution history so Maia AI Agents can reason about what's fresh, what's stale, and how data actually flows.
• Scheduling: Run crawlers on a standard interval or an advanced cron schedule, with pause/resume controls.
• Access controls: Choose Public (every project) or Restricted (project allowlist) per graph, and manage who can view or manage each graph at the user level.

Read the docs to get started.

🎯 Mission Control is now in public preview

We're excited to introduce Mission Control—a centralized hub to launch, monitor, and review the work your Maia AI Agents are doing for you. Available now in public preview, Mission Control also introduces two new capabilities: multi-modal chat and “bypass permissions” mode.

Until now, managing Maia work meant staying within a single conversation. Mission Control transforms this experience by providing a kanban-board-level view of everything Maia AI Agents are doing across your projects, enabling you to run multiple tasks in parallel and only intervene when needed.

Key features:

• Kanban board: Every task flows through Backlog → In Progress → Needs Attention → Completed for at-a-glance status tracking.
• Quick task launch: Select a project, branch (or create a fresh isolated working branch), environment, and optional knowledge graph, then describe your requirements.
• Auto-start capability: Tasks begin processing automatically when created—no manual intervention required.
• Live conversation panel: Monitor real-time progress and send follow-up messages without leaving the board.
• Needs Attention column: Tasks requiring your review are clearly surfaced in their own dedicated space.
• Seamless review workflow: Jump to Review in Designer to inspect changes, then move to completed when satisfied.
• Advanced filtering: Sort by project, source, and time period, or search across all tasks.

📎 Multi-modal chat

You can now attach images and PDFs directly to your task prompt in Mission Control. Give Maia AI Agents a supporting diagram of the pipeline you want built, a screenshot of an error you're hitting, or a PDF containing the spec from your team—and they can work straight from the source instead of from your description of it.

• Attach images: Drop in a supporting diagram, screenshot, mockup, or whiteboard photo—perfect for pipeline diagrams or system designs you'd otherwise have to describe in words.
• Attach PDFs: Share a spec, requirements doc, or report you want Maia AI Agents to use as context for the task.

Use the attach (+) control in the prompt area to add an image or PDF to your message.

🔓 Bypass permissions mode

Bypass permissions mode is the opposite of Ask permission mode. Tool calls execute immediately, with no approvals, decisions, or check-ins—great for trusted, hands-off runs like prototyping or clearing a backlog in a sandbox. Under the hood it's the same as approving every permission prompt as it comes up; Maia AI Agents aren't unlocking new capabilities, only skipping the checkpoints.

We don't recommend leaving Bypass permissions on by default. Ask permission should be your normal mode. Flip to Bypass for specific tasks where you know what Maia AI Agents are doing and you've scoped the environment for it. Use it in safe, well-scoped environments:

• A disposable project or branch where unintended changes are easy to throw away (scratch branch, dev environment, sandbox project)
• A warehouse user/role with only the access Maia AI Agents need. Point them at a dev warehouse or a least-privilege role, not your production admin credentials
• Approval gates on when Maia AI Agents are working from untrusted input like a PDF or image you didn't author

Mission Control is currently user-scoped, meaning you'll only see your own tasks.

You can access Mission Control through the AI Agents icon in the left navigation. Click + Add task to launch your first task. Any tasks requiring your input will appear in the Needs Attention column.

Read the docs to get started.

💬 We'd love to hear from you!

Let us know how these new features are improving your workflows—we're all ears! Feel free to add any comments or questions below.

Want to get involved?

Join the Matillion Community to stay up to date, share feedback, and help shape our product roadmap for future innovations.

‍