The Vibe-Code Agent Builder
Build an agent by chatting
You'll be able to
- Describe a goal in plain language and have Glean draft an agent from it
- Read the generated plan and decide whether it matches your intent
Read first
Day 1 Lab
**Status: Generally available — Glean’s Agent Builder and its core features (scheduled triggers, versioning, native actions) are GA. Only the *content triggers* trigger type remains Beta; confirm availability in your tenant.**
- Pick one small, repeatable chore you do by hand — "summarize new #incidents posts each morning," say.
- In the agent builder, describe that goal in one or two plain sentences. Don't configure anything yet.
- Read the draft the builder proposes back: its named steps and the apps it wants to touch.
- Compare the draft to what you actually meant, and note the first place it diverges.
- Reply in chat to correct that one divergence — then watch the plan redraw.
Working Example: From a Sentence to a Draft Plan
Per Glean's Fall '25 launch announcement, you build an agent conversationally: you state the goal, the builder drafts a plan, and you read it before committing. Verified against fall25_launch.
You: "Each morning, summarize new posts in #incidents and DM me the digest."
Builder draft (proposed, not yet running):
Step 1 Read new messages in #incidents since last run
Step 2 Summarize them into a short digest
Step 3 Send the digest to you via DM
You read the draft and ask: which channel? whose DM? what counts as "new"?
The plan is a starting point you steer — not a finished agent you trust on sight.The vibe-code flow moves the work from filling out forms to reading and correcting a draft. That's faster, but it means the first real skill is *reviewing the plan* — the builder's guess about your intent is exactly where MAP-style scrutiny belongs. Treat this as an announced capability, not a settled one: confirm the builder exists and behaves this way in your own tenant before you lean on it.
Check for understanding
After you describe a goal, what is the first thing you should check in the draft before trusting it?
Check yourself
1. In Glean's vibe-code agent builder, how do you start building an agent?
2. The vibe-code builder, 100+ native actions, scheduled triggers, looping, and versioning are best described as...
3. Before adding a native WRITE action (e.g., update a Salesforce case) to an agent step, what should you check?
4. An agent runs unattended on a daily schedule. Which guard most directly keeps a bad run from compounding?
5. Why does rolling an agent back to a known-good version count as a governance control?
In 10 seconds
“Glean's Fall '25 launch pitched a vibe-code builder: describe what you want in plain language and get a draft agent back. You read its plan, then refine — chatting, not configuring forms.”
Agent Steps and Instructions