Tray Merlin Agent Builder
Move AI agents from experiments into production
Most teams can build an agent that answers questions. Fewer can run agents that interact with tools, follow guardrails, and scale across teams. Tray helps teams do that.
How Apollo runs AI agents in production
Apollo built an ITSM agent on Tray Merlin Agent Builder that now deflects nearly 40% of IT tickets. The agent:
- Lives in Slack, where IT requests already happen
- Takes action through connected tools, including password resets and license access
- Uses guardrails, starting with responses only when confident
- Escalates cleanly when a human needs to step in
The result is faster resolution without degrading CSAT.
The gap between agent pilots and production
Teams don’t fail because they can’t build agents. They stall because running agents across teams introduces problems most tools don’t solve.
Where agent efforts break down
- Agents answer questions but can’t complete tasks
- Tool access is handled case at a time
- Logic is rebuilt for every new agent
- Guardrails are added after rollout
- Each team works differently
This slows adoption and increases risk as usage grows.
What changes with Tray
- Agents interact with tools through governed access
- Teams reuse approved components instead of rebuilding
- One approach teams can share and extend
- Guardrails are defined before rollout
- New agents don’t mean new patterns
This makes it possible to scale agents without starting over.
Talk through your agent use case
Get a walkthrough focused on how teams build and extend agents using governed tools and reusable components.