Skip to main content
Each time you chat with in or , you start a new session. A session is an ongoing, structured interaction between you and , which includes your full conversation history, relevant context, and any tool based actions have performed on your behalf in your branch or chosen branch in . Sessions are:
  • Context-aware: build on prior messages within the same session, allowing for more coherent, informed responses across multiple prompts.
  • Session-scoped: In , your chat with remains active and visible in the panel as long as you’re working in the same project and branch. Once you close a branch, your conversation is saved in your conversation history.
  • Private: Your conversations with are only visible to you. When you open a branch in , or when viewing tasks in , the only conversations and tasks you can see are your own.
In , you can switch between pipelines in the same branch without losing session context. However, switching branches will close your session and save your conversation in the conversation history.

Chatting with Maia AI Agents

For the best results, we recommend starting a new conversation or task for each piece of work you want to assist with. This helps to provide more focused, relevant assistance. sometimes need more context to get things right, particularly when planning pipelines or working on complex prompts. When need more information from you, they will present clear, structured questions to help clarify how to respond to your prompts. These can be:
  • Multiple choice questions: Choose from predefined options, or provide your own answer with the “Other” option.
  • Multi-step forms: When need to ask several questions, they will present them one at a time with step-by-step navigation.
In , if appear to misunderstand your intent or lose context, click New Chat in the top right of the chat panel to start a new conversation. If you have a long conversation with , they may summarize your chat so far, and then continue.

Visualizing data

In chats and tasks, can visualize the data they sample in a component directly in the chat panel. These visualizations enable you to discover key insights and identify trends. To visualize your data in the panel, prompt to create a chart or graph representing the data in a specific component. These visualizations are based on sampled data and don’t include the entire dataset. For more information about how sample data, read Sampling data.

Plan mode

Use plan mode in and to chat with about your work without making any changes. Plan mode provides you with extra visibility and control over ’ suggested actions, which can be useful when working on complex pipelines, and enables you to suggest changes to ’ approach before any work is performed on your pipelines. In plan mode, will analyze your prompt, explore your project and data warehouse, then present a multi-step plan to achieve your goal without making any changes to your pipeline. You can then suggest changes to the plan if needed, or accept ’ suggestion. will only make changes to your pipeline after you have approved the suggested plan. To use plan mode, toggle on Plan mode in the dialog when creating a task in , or at the bottom of the chat panel at any time when working on a task in or . will automatically suggest switching to plan mode if they think your prompt is complex enough to require planning. When this happens, you will see a message in the chat panel—click Accept to allow to use plan mode. Expand this box to watch our video about using plan mode for complex tasks.


Conversation history

In Designer, Maia AI Agents automatically save your conversations, making it easy to continue where you left off or revisit previous work. This ensures that important context and progress aren’t lost when you need to step away or switch tasks. Your conversation history with Maia AI Agents is kept for two years. None of your data is used to train third-party AI models or to enhance Matillion’s AI services. Your conversation history is unique to your current branch. Switching to a different branch, e.g. from main to developer-1, will show your conversations on that branch. You can never view another user’s conversations with Maia AI Agents, even if they worked on the same branch as you. The 10 most recent conversations on your branch are available for you to resume, rename or delete.

Resuming recent conversations

When you open the chat panel, the Recent section shows the titles of your most recent conversations. dynamically update the title of your conversation as you work, so that it always reflects the topic you’re discussing. Click a conversation to open it and resume your chat with . When you select a previous conversation, load the complete chat history and you can continue exactly where you left off. To see more conversations, either click See all above the recent conversations, or click Chat history in the top right of the chat panel.

Renaming and deleting conversations

You can rename and delete your previous conversations with . Renaming a conversation makes it easier to find useful pipeline information or data analysis, while deleting conversations lets you clear conversations you no longer need. If the conversation you want to rename or delete is not shown in the Recent section, click Chat history in the top right of the chat panel to see all your conversations. To rename or delete a previous conversation, hover over the conversation tile, and click the three dots that appear. Next, click Rename or Delete.

Explore files

can search through and explore files in your project, making it easier to understand existing work, find specific configurations, or perform project-wide updates. ’ file search capabilities help you:
  • Locate specific files: Find pipeline files, scripts, or documentation by name or pattern.
  • Discover existing components: Find configured components to copy or reference.
  • Validate project-wide changes: Ensure parameters or configurations are updated everywhere they’re used.
  • Understand your project structure: Get an overview of what files and pipelines exist in your project.
For example, you could prompt as follows:
  • "Find all transformation pipelines that connect to the "product-orders" table."
  • "Show me any existing Salesforce load components I can copy."
  • "Find every time the database name 'PROD_DB' is mentioned."
  • "Summarize the content of the pipeline files in the analytics folder."

Search performance

For larger projects with many files, searches may take longer to complete. will let you know when they’re working on your search request. Using file search is much more efficient than having view each file individually, especially when you need to:
  • Perform large-scale refactors across multiple pipelines.
  • Check whether changes were applied consistently throughout your project.
  • Find specific patterns or configurations across many files.

Slash commands

In , slash commands are a command-line interface (CLI) style method of interacting with . These commands simplify complex prompts into a one-word command to standardize common tasks. The table below explains what will do as a response to each slash command. To use a slash command, type a forward slash / in the chat panel, followed by the command you want to give. After you type /, an autocomplete list of available slash commands will appear. After typing the command, enter the name of the object that you want to process—you can use the @ symbol here to mention a specific file in your project. For example, you could prompt “/document the customer_data folder”, “/explain this pipeline” or “/fix @weekly_orders”.
Command’ task
/documentGenerate a Markdown file containing information about a specified pipeline, file or directory. The information provide will include a summary of the file, the components used (when documenting a pipeline), how to use the file, and important information like dependencies and performance considerations.
/explainGenerate a brief explanation of a specified pipeline or concept. The information provide will include a short explanation of the pipeline’s purpose, data flow, and key components.
/fixIdentify issues in a specified pipeline and suggest ways to fix them. will determine the cause of any errors or configuration problems in the pipeline and present you with options to resolve the issues. will not take any action until you approve one of the options presented.
/multi-table-loadLoad multiple tables from a data source using incremental or full load.
/notesAdd notes to a specified pipeline. will add notes to the canvas with information about the purpose of components, the data flow in the pipeline, any potential issues, and opportunities for optimization.
/optimizeAnalyze a specified pipeline and suggest opportunities for optimization. The suggestions provide can include information about generic SQL that could use specialized components, opportunities for parallelization, and performance bottlenecks. will not make any changes to your pipeline until you approve one or more of their suggestions.
Expand this box to watch our video about using slash commands when chatting to .


Tool permissions

use tools to perform tasks in your workspace based on your prompts. These tools let go beyond chat and perform real, interactive actions. tools allow them to create and modify pipelines, add and configure components, search your data warehouse, preview or sample data, and run and monitor pipelines. For security and transparency, ask you for permission before using tools that can affect your work. When relevant, you will see a permissions dialog in the chat panel. This permission-based model ensures that you stay in control while enabling to act confidently on your behalf. You can:
  • Accept: Approve the action for a one-time use.
  • Accept for session: Allow to repeat this action without asking again during the current session.
  • Decline: Deny the request and optionally provide a reason for declining. This helps adjust their response or better understand your prompt.

Background notifications

Background notifications enable to notify you when they finish a task or require your input, such as requiring permission to sample data. This means you can give a prompt, then switch to a different browser tab or program while are working, and resume your task as soon as finish. Background notifications are provided as:
  • Tab indicators: A dot in the browser tab title.
  • Audio chimes: A notification sound.
  • Browser notifications: A system-level notification, if you grant permission to show browser notifications.
To activate background notifications:
  1. Open the chat panel in .
  2. Click the three dots in the top right to open the Settings drop-down menu.
  3. Toggle on Notifications.

Test suggestions

In , can suggest and create test pipelines when building or modifying pipelines. When you update a pipeline, will also automatically flag any existing tests that are affected. To enable to work with test pipelines:
  1. Open the chat panel in .
  2. Click the three dots in the top right to open the Settings drop-down menu.
  3. Toggle on Test suggestions.

Canceling a prompt

If you want to cancel ’ current response, click Cancel—this icon appears in the input field of the chat panel when are responding to a prompt. This stops the current response but keeps your session open. You can continue the conversation or ask something new immediately. Canceling a prompt does not clear the conversation history.