When your agents work, they produce outputs. A research task yields a summary. A browser session extracts data from web pages. A computer session generates files. These outputs are collectively called deliverables—the tangible results of your agent's work that you can access, download, and use.
O-mega's deliverable system ensures that valuable outputs don't disappear into chat history. Every meaningful result your agent produces is captured, categorized, and made available in organized form. This means you can always go back and find what your agents have created.
What Counts as a Deliverable
Deliverables encompass all the useful outputs your agents produce during their work. The system is designed to capture everything of value without requiring you to explicitly save or tag things.
Types of deliverables include:
- Extracted data - Text, tables, and lists pulled from web pages during browser sessions
- Generated files - Documents, spreadsheets, code files, and websites created in computer sessions
- Downloaded files - PDFs, images, and exported data that agents download from websites
- Screenshots - Visual captures of browser state at important moments
- Summaries - Synthesized information from research tasks and complex workflows
The key insight is that deliverables are more than just files. They include structured data extractions, summaries, and any other meaningful output—not just things that exist as separate files.
How Deliverables Are Created
Deliverables are extracted automatically as your agents work. You don't need to explicitly request that something be saved—the system recognizes valuable outputs and captures them.
The extraction process happens throughout task execution:
- Step-by-step extraction - During browser and computer sessions, important findings are captured at each step
- File downloads - Any files your agent downloads from websites are automatically saved
- Generated outputs - Files created in computer sessions (websites, documents, processed data) are captured
- Final summaries - When tasks complete, aggregated results are stored as deliverables
Each deliverable is linked to the session and step that produced it. This provenance tracking lets you trace back to exactly when and how a deliverable was created if you need to understand its context.
Accessing Deliverables
Deliverables surface in multiple places throughout O-mega, making them easy to find regardless of where you are in the interface.
Where to find deliverables:
- Chat - Deliverables often appear inline in your conversation, right alongside the agent's response about them
- Files tab - All files generated by or for an agent are collected in the Files section
- Task history - Each task execution shows the deliverables it produced
- Browser session logs - Step-by-step deliverables from browser automation are visible in session details
The multiple access points mean you can find what you need from wherever you happen to be working.
Deliverable Types
Different kinds of work produce different kinds of deliverables. The system categorizes them to help with organization and filtering.
| Type | Description |
|---|---|
| Text | Extracted or generated text content, including summaries |
| File | Downloaded or created files of any type |
| Data | Structured data like JSON objects or table extractions |
| Screenshot | Visual captures of browser state or generated output |
This categorization helps when you're looking for something specific. If you know you need the spreadsheet from last week's analysis, you can filter for files rather than scrolling through text summaries.
Deduplication
When agents work on related tasks, they sometimes extract the same information multiple times. Rather than creating duplicate entries, O-mega's deliverable system deduplicates automatically.
If identical content is extracted at different steps, it's stored once with a reference count. This keeps your deliverable history clean and avoids cluttering your Files section with redundant copies of the same information.
Using Deliverables
Once deliverables are saved, they become part of your ongoing workspace. They're not just archives—they're active resources you can work with.
What you can do with deliverables:
- Download files directly to your computer for use in other applications
- Copy text content to your clipboard for pasting elsewhere
- Reference deliverables in future conversations with your agent
- Use files as inputs for subsequent tasks
Your agent is aware of deliverables it has created. You can ask "use the competitor analysis from last week" and your agent will know what you mean, pulling in that deliverable as context for new work.
Related: Creating Tasks | Computer Sessions Overview