When you need a competitive landscape across 20 pipeline assets, a cross-trial comparison for an upcoming advisory board, or a structured screen before a BD meeting, you can delegate the entire analysis to the AI Research Agent. Describe your request, review the execution plan, and receive finished deliverables: structured tables, cited reports, and supporting evidence.

From the Home Screen, locate Assign a Task to the Research Agent. Enter your research question along with any context the agent should consider. You can optionally select a workflow type to guide the analysis structure, or tag a workflow directly in the input field.

Plan Mode lets you review the agent's approach before execution begins. After you submit your query, the agent outlines the research steps and any deliverables it will produce (tables, reports, models). You can edit individual steps or approve the full plan.

If your query does not require deliverables, the plan reflects this and the agent returns an analysis summary only.

Auto Mode skips the plan review. The agent determines the approach and proceeds directly to execution, producing the same structured outputs.

In Plan Mode, the agent presents a structured list of research steps before doing any work. You can add new steps with Add Step, edit existing ones, or remove steps that are not relevant to your analysis.

When you are satisfied with the plan as presented, click Accept & Execute Plan to begin. If you want to adjust the approach first, click Provide Feedback to send the agent a follow-up instruction.

If you edit any steps directly, the action buttons update to reflect your changes:

Once execution starts, the agent works through each research step and shows its progress in real time. You can click Stop Execution at any point if you need to revise the approach.

You do not need to stay on the Assistant tab while the agent runs. Navigate to other parts of the platform, and completed deliverables (tables, reports) will appear in your workspace folder when finished.

When the agent completes your request, the response includes the full chain of reasoning alongside the deliverables. You will see the research steps the agent followed, a summary consolidating key findings, and any tables, reports, or figures produced during execution, all embedded directly in the response.

This transparency lets you trace every conclusion back to the data the agent retrieved, so you can validate findings before incorporating them into your work.

Workflows are repeatable templates that define methodology, expected outputs, and quality criteria for specific analysis types (landscapes, cross-trial comparisons, asset deep dives, and more). They ensure consistency across your team's analyses.

Type / in the input field and choose from the dropdown, or click one of the quick-access workflows below the text box. Available workflows include Maven Bio built-in templates, organization-wide workflows your team has published, and custom personal workflows.

To create your own:

Provide additional context by attaching existing Maven Bio artifacts or uploading files. Click the attach icon, search for a previously created table, report, or entity, and select it from the dropdown.

You can also upload supporting files (PDF, DOCX, PPTX, CSV, XLSX, TXT, MD, PNG, JPG, JPEG, WEBP, GIF) up to 25 MB each, with a combined limit of five attachments.

Type @ followed by an entity name to tag drugs, companies, indications, targets, mechanisms, reports, tables, or blueprints directly in your query. A dropdown appears with matching results. Selecting an entity inserts it as a structured tag, helping the agent focus its research precisely.

Once you receive your analysis, you can refine or build on the results: