When you need a fast, cited answer mid-analysis (the latest FDA action on a competitor's asset, a quick mechanism comparison, or a summary of recent trial readouts), the Assistant's Chat mode delivers sourced responses without leaving your current workflow. For structured, multi-step research that produces Smart Tables, reports, or spreadsheets, switch to Agent mode.

The Assistant supports two modes, accessible from the central input bar on the start page and the sidebar throughout the platform:

Chat is for conversational queries: quick lookups, summaries, comparisons, and follow-up questions. Responses are cited and appear directly in the conversation.

Agent is for structured research tasks that produce deliverables (Smart Tables, reports, spreadsheets, slide decks). The Agent plans a multi-step workflow, executes it, and returns organized outputs.

If you ask a question in Chat that would benefit from structured analysis, the Assistant will suggest switching to Agent mode. You can continue in Chat or switch with one click.

To direct the Assistant toward specific entities or your existing work, use @mentions inline. Type @ followed by a category (drug, company, indication, target, mechanism, report, table, or blueprint), then press space and start typing the name. A dropdown appears with matching items from Maven Bio's database and your workspace. Select the item to insert it as a tagged reference in your message.

For example, typing @drug sotorasib ensures the Assistant grounds its research on the correct asset, avoiding ambiguity when names overlap across indications or companies.

To provide additional context, click Attach to open the attachment panel. You can select existing Smart Tables, reports, or Maven Ontology items from the left panel, or upload files from your device. Supported formats include PDF, DOCX, PPTX, CSV, XLSX, TXT, and MD, with a 2.5 MB limit per file.

You can combine up to five @mentions and file attachments total per message.

Before submitting your query, select a reasoning mode to match the depth you need:

Fast is optimized for speed and surface-level retrieval. Use it for fact checks, quick drug lookups, or pulling trial phase data. Example: "List PDE-5 inhibitors currently in clinical development."

Advanced balances speed with moderate analytical depth. Use it for comparisons, structured summaries, or trial design reviews. Example: "Compare the efficacy endpoints of Wegovy and Zepbound Phase 3 trials."

Max prioritizes depth for logic-heavy, nuanced queries. Use it for cross-study interpretation, hypothesis generation, or mechanistic analysis. Example: "What differentiates PDE-5 inhibitors mechanistically and clinically across indications?"

Chat responses include inline source citations with direct links for verification. The Assistant highlights key findings and ranks sources by relevance. If a response includes tabular data, you can download it as CSV or copy it for use elsewhere.

To start a new conversation, press Shift + C or click the + icon in the toolbar. To revisit past conversations, press Shift + H or click the history icon in the assistant toolbar. This opens a list of recent conversations in your current workspace, with an option to View all chats across workspaces.

The full chat history view includes both Chat and Agent conversations. Each card shows the workspace, date, and message count. Use the search bar and date filters to find any past analysis.