Skip to content
RebelCore Docs

RebelCore™ Agent

RebelCore Agent lets you ask questions about semantic datasets you are authorised to use. Open it from a ready model in the Tree, from an authorised Agent launcher, or at your organisation’s Agent address.

Agent-only users go directly to Agent after signing in.

Start a session

  1. Click NEW SESSION.
  2. Open the settings cog.
  3. Select one or more datasets.
  4. Choose Fast or Reasoning.
  5. Review Mask to AI and Mask on interface.
  6. Type a focused question in Ask something….
  7. Click the send icon or press Enter.

The response appears in the conversation. Ask follow-up questions in the same session when they relate to the same task.

Choose datasets

Open the settings cog to see the semantic datasets available to you. Tick the datasets relevant to the question.

Each dataset shows its LLM exposure level. When several datasets are selected, Agent displays the effective level for the chat. The most restrictive selected level applies.

Only datasets from projects you can access are listed. If project access changes, a dataset may disappear from your selection.

Fast and Reasoning

ModeBest forExperience
FastDirect questions and everyday analysis.Streams progress and the answer as it becomes available.
ReasoningMore involved questions that may require several data actions.Provides an expanded activity log where included in your licence.

The mode is fixed after the first prompt in a session. Start a new session to change it.

Reasoning availability can depend on your licence. If the option is disabled, use Fast or contact your Customer Super Admin.

Privacy controls

The settings cog contains two privacy controls.

Mask to AI

When enabled, recognised personal identifiers are replaced before AI processing. This can include:

  • Names.
  • Email addresses and telephone numbers.
  • Account and customer identifiers.
  • National, tax, health, and social-security identifiers used in the United States, United Kingdom, European Union, Australia, and South Africa.

Amounts, dates, percentages, roles, departments, and other useful business context remain available unless they are part of a personal identifier.

Mask to AI is fixed after the first prompt. Start a new session to change it.

Your Customer Super Admin can require Mask to AI for the organisation. When required, the setting remains on and cannot be changed by the Agent user.

Mask on interface

When enabled, recognised personal information remains masked in the conversation you see. When disabled, Agent can display authorised original values that are available for the session.

You can change Mask on interface during a session because it does not change what was already sent for AI processing.

RebelVault Files

Use RebelVault to upload TXT, PDF, DOCX, XLSX, XLS, CSV, or PPTX documents that Agent can search when answering your questions.

RebelVault is controlled separately from selected semantic datasets. Open the settings cog and use Allow File Content to decide whether uploaded file content can be searched for the next prompt.

Exposure badges

Each Agent answer displays Full, Limited, or Advisory:

  • Full can include exact authorised data.
  • Limited provides analysis without source records.
  • Advisory provides qualitative guidance without dataset values or statistics.

The badge describes the output policy for that response. It is separate from personal-information masking.

Ask effective questions

Include:

  • The person, account, period, region, or category you mean.
  • The measure or outcome required.
  • The format you want, such as a short summary or table.
  • Any comparison or filter that matters.

For example:

Show the salary, tax deduction, and take-home pay for the selected Sales employee for the May 2026 payment.

If Agent finds several matches, add a department, role, date, or another authorised identifier.

Sessions and audit

  • Sessions on the left contains your saved conversations.
  • AUDIT on the right shows the available progress and activity for the current conversation.
  • Customer Super Admins can use the portal Audit page for organisation-wide review.

Limits

  • Agent can only use ready datasets you are authorised to access.
  • Output detail remains subject to the effective exposure level.
  • Destructive administration remains in the customer portal.
  • A session under human review can be temporarily paused or ended.