Unify Logo Footer.svg
AI Governance
Logo
Discovery Troubleshooting & FAQ

Discovery Troubleshooting & FAQ

Logo

3 mins READ

Introduction

Most Discovery problems are one of three things: the credential cannot see what you expect, the sweep has not run yet, or the asset is there and you are looking in the wrong list. This page is ordered by how often each turns out to be the answer.

A connection finds nothing

  1. Has the first sweep finished? — A new connection shows "Initial sync in progress" and hides its header actions until it completes. Large sources take a while.

  2. Can the credential see the assets? — Discovery reads what its connection is permitted to read, and nothing more. This is the most common cause by a wide margin — the connection works, authenticates, and is scoped to a subset you did not intend.

  3. Do the assets exist where you think? — Different account, different region, different workspace in the source platform.

  4. Force a sweep — Sync Now separates a scheduling problem from a permissions one. If an immediate sweep also finds nothing, it is not the schedule.

One specific agent never appears

Check

Because

Is it excluded?

Excluded assets are filtered out of the Agent Registry entirely. Look on the connection's own detail page

Is it a model?

Discovery returns agents and models together, but they split into two registries afterwards

Is it in another workspace?

Workspace is a visibility boundary, not a label

Was it created after the last sweep?

Check the sync interval, then Sync Now

Note: "The agent has disappeared" is almost always exclusion. Excluding an asset keeps it in Discovery and removes it from the registry. Nothing is lost — it is one screen away, on the connection it came from. See Review & Triage Discovered Assets.

Sync fails

"Sync failed" with Try again means the sweep ran and failed — read the error shown, fix the connection, then Try again. In order of likelihood, the cause is: expired or rotated credentials, a permission revoked on the source side, or the source being unreachable from UnifyApps.

A connection that hits a sync error deliberately leaves the syncing state, so the page does not spin forever — and so its settings stay reachable while you fix the cause.

Frequently asked

  1. Does discovering an agent govern it? — No. Discovery finds; registration governs. They are separate acts, and the separation is deliberate — see Working with Discovered Agents.

  2. Does a sweep cost anything? — A sweep reads the source's inventory. It does not run your agents and does not consume model tokens.

  3. Can two connections point at the same source? — Yes, and sometimes they should — two credentials with different scopes are genuinely two views. Expect overlap in what they return.

  4. Why did the discovered banner stop appearing? — Dismissal is remembered per connection in your own browser. The assets are still there; open the connection's asset list.

  5. Can I change a connection's tier after creating it? — The tier depends on what the platform's integration supports. See Connection Types.

  6. What happens to excluded assets on the next sweep? — They stay excluded. Exclusion is a decision on the record, not a filter that resets.

  7. Should I turn on auto-register? — It depends on whether the source contains only production assets or a long tail of experiments — see Sync Cadence & Auto-Register.

If none of this helps

Bring three things to your UnifyApps account team, and the question usually resolves in one exchange: which connection, what you expected to see, and whether Sync Now changed anything. The third answer is the one that splits the problem in half.