Is there a better way to surface custom agents? [Champions Slack Insider]

[#1]

This one is less about features… and more about reality from @Paulo Ramalho:

“We built great Rovo agents—but no one uses them.”

If you’re working with large organizations, this probably sounds familiar.

The real problem isn’t building agents—it’s finding them

In theory:

• Users can switch agents
• Search for them
• Bookmark favorites

In practice:

• Too many clicks
• Too much remembering
• Too little context

So what happens?

• Users default to generic Rovo chat
• Custom agents get ignored

Why this happens (especially at scale)

In large orgs (5,000+ users):

• Agent lists keep growing
• Naming isn’t always consistent
• Users don’t know which agent to use
• There’s no strong in-context surfacing

Discovery becomes a memory problem, not a UX feature

What about Forge or APIs?

Paulo tried the logical path:

• Build a Forge app
• Surface agents directly in context (e.g., issue view)

Blocked by this limitation- Forge can only open agents created within the same app. Meaning, studio-built agents ≠ accessible via Forge bridge and today there is no supported API to list or surface all agents.

Also important: No short-term (1–2 month) roadmap commitment for this yet.

So what actually works today?

Not perfect—but effective.

Option 1: Confluence as an “Agent Catalog”

Yes, it sounds basic—but it works when done right. Instead of a random list of agents, create a structured table with:

• Agent name
• Purpose
• When to use it
• Example prompts
• Where it works (Jira, Confluence, etc.)

Option 2: The “Agent to Find Agents” pattern

This was the breakthrough idea- Build a Rovo Adoption Agent

What it does:

• Reads your agent catalog (Confluence page)
• Asks 1–3 clarifying questions
• Recommends the best agent
• Provides:
  • Agent name
  • When to use it
  • Starter prompt

Why this works

It removes the hardest part: “Which agent should I use?”

Instead of:

• Searching
• Remembering
• Guessing

Users just describe their goal.

Important implementation tips

Keep it tight:

• Limit catalog to top 10–20 agents
• Use consistent naming
• Include real example prompts
• Define clear “when to use” scenarios

Too many agents = same problem, different format

If you’re waiting for the perfect UI to solve this… You’ll be waiting a while. If you design for how users actually behave today, you’ll see adoption move.

[Rebekka Heilmann _viadee_]

Most importantly, have a good Usage Policy that tells people to:

• use comprehensive names and descriptions
• Check for duplicate Agents
• and most importantly: limit Agents to the Users they are relevant for

I'd also recommend

• a Sandbox environment where people can test and an approval process for the move to production
• a "hoop" to jump in order to get Agent Creation permissions

Got an article cooking on the details of all this :)