Skip to main content

Documentation Index

Fetch the complete documentation index at: https://slackhive.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

Each agent can have one or more audiences — named user groups that the agent treats differently. Use audiences when the same agent serves multiple teams who need different tone, depth, or focus. Audience configuration panel showing name, priority, instructions, and member list

How it works

Each audience entry defines:
  • A name for the group (e.g. “Engineering”, “Sales”, “Support”)
  • A priority — lower number applies first when a user belongs to multiple audiences
  • An instructions block that is appended to the agent’s base system prompt for members of this group — written by you and polished by AI before saving
  • An optional verbose override — overrides the agent’s default verbose setting for members of this audience
When a Slack user messages the agent, SlackHive checks which audience (if any) they belong to and applies the matching prompt override on top of the base system prompt.

Setting up audiences

  1. Open the agent → Instructions tab → Audience section
  2. Click New Audience
  3. Give the audience a name and priority
  4. Write the instructions (tone, focus, constraints for this group)
  5. Click Polish with AI to have Claude refine the wording, then Save
  6. Add members from the right-hand panel

Priority

If a user belongs to multiple audiences, the one with the lowest priority number wins. Use this to create a catch-all audience at a high number (e.g. 100) and more specific overrides at lower numbers (e.g. 10).

Access control

Configuring an agent’s audiences requires editor access to that specific agent — a global editor role alone is not sufficient. Admins and superadmins can always manage audiences for any agent.

Verbose override

Each audience has its own verbose toggle. When set, it overrides the agent-level verbose setting for members of that audience. This lets you give one group streaming step-by-step output and another group clean final-answer-only responses from the same agent.