The Vela agent workspace
Chat with Vela or a specialist, attach files and follow work in one place.
Vela is the personal name of your Operator: the agent that coordinates Search, Insights, Engage and Optimise. Together they form your five-agent team.
The workspace opens as a separate rounded window above the current page. It can close or minimise without rearranging the page. Click a character to talk to that agent directly.
| Part | What it does |
|---|---|
| Agent list | Selects Vela or one of the four specialists. |
| Vela's team | Opens the team's recorded work. |
| Conversation header | Names the agent you are talking to. |
| Message box | Accepts questions and instructions directly. |
| Attach | Adds documents or images to your message. |
| Dictate | Turns speech into editable text. |
| Talk | Starts a spoken conversation using ElevenLabs voice. |
| Details | Shows the agent's role, routines and recorded progress. |
| Library | Shows available uploaded files and recorded results. |
Selecting Search opens Search's conversation. Its replies and thinking indicator use Search's own character and name. Vela's replies use Vela's character. Your unsent draft stays with its conversation. There is no answer-mode selector or separate task form: describe the outcome in chat.
Progress and results
Vela can answer directly or delegate a bounded task to a specialist. The conversation shows real job states and available results. Background results return to their original conversation and Vela's linked conversation. Source permissions still apply, including when you return later. Work that is queued or waiting for review is not shown as completed.
Attach a file
Use Attach and add an instruction, such as "Read this PDF and identify the company and next steps." Attach up to three files per message, up to 8 MB each. Supported files include PDF, Word DOCX, images, plain text and EML email files. Older .doc files need conversion to .docx or PDF.
Uploads remain private to your signed-in account and workspace. Extraction is bounded: PDFs use text from the first six pages and images of the first two pages; Word extraction and embedded images also have limits. The assistant should state omissions rather than imply it read an entire long file. Scanned pages can be read visually within these limits. Library keeps available uploads and results; it is not a remote computer.
Ask in chat to set up a recurring task. Routines and their next run appear under Details. See Get things done in Vela.
Review when an action needs it
Protected changes show their actual target and content before approval. Sending, publishing, campaign starts and paid actions retain their application checks. A draft is not a sent message. See Approval cards and connected assistant reviews.
The public website version explains the product and shows labelled examples. It cannot access your private files, customer records or live account tasks.