Provisioning accounts in your apps
Keep the accounts in an app your organization registers in step with who is assigned to it, over SCIM 2.0 or through Microsoft Entra, with every attempt and every difference on record.
An app that signs people in through Scrydon often keeps its own account for each of them. Provisioning keeps those accounts in step for you. When someone is assigned to the app, Scrydon creates their account there. When their details or teams change, it updates the account. When they leave the organization, are suspended, or are no longer assigned, it deactivates the account. Changes reach the app within minutes.
Provisioning is set on the app, under Settings → Organization → Apps → your app → Provisioning. It works the same for an OIDC app and a SAML app. Only an owner or an admin of the organization can see or change it. Every change is recorded in the audit log.
Who is provisioned
The people provisioned are the people allowed to use the app: every member when the app admits all members, or the teams and people you named when it is restricted. Narrowing who can use the app deactivates the accounts of the people it no longer admits.
Two ways to reach the app
Each app has one provisioning target, reached one of two ways.
| Way | Use it when | What you enter |
|---|---|---|
| SCIM 2.0 | The app has a SCIM 2.0 endpoint. | The endpoint's base URL, and the organization secret that holds its bearer token. |
| Microsoft Entra | The app's accounts are kept by an Entra enterprise application. | The organization's Microsoft connection, the enterprise application's service principal id, and its provisioning job id. |
The base URL must use https, with no user name, password, query or fragment, and its host must be on your organization's egress allowlist. Scrydon refuses any other address.
You never type a token into the form. You choose an organization secret, or an extension connection, that already holds it. Scrydon stores which one you chose, never its value. To change the token, update the secret: the target uses the new value from its next call. Create the organization secret before you set up the target.
What a SCIM 2.0 app receives
Each person is a SCIM User:
| Attribute | Value |
|---|---|
userName | The person's sign-in name in your organization (their email when they have none). |
externalId | A fixed identifier for the person that never changes. Match accounts on it. |
displayName, name.givenName, name.familyName | The person's name. |
emails | One primary work email. |
active | false once the person is deactivated. |
A new person is created with POST /Users. Every later change is a PATCH.
Deactivating an account sets active to false: Scrydon never deletes an
account, so its data stays in the app. Each team the person is in becomes a
SCIM Group whose externalId is the team's identifier, and the person is
added to it or removed from it as their teams change. Groups the app created
itself are never changed.
When the app already has an account with the same userName and the same
externalId, Scrydon takes that account over instead of creating a second
one. An account with the same userName and a different externalId is left
alone, and the attempt fails.
Set up a target
- Open the app and choose Provisioning.
- Choose Set up provisioning.
- Pick SCIM 2.0 or Microsoft Entra, then fill in the fields above.
- Choose Save target. A new target starts off.
- Choose Turn on when you are ready. Scrydon then sends every assigned person, and keeps them in step from then on.
The way a target reaches the app is fixed once it is saved. To switch from SCIM to Entra, turn the target off, delete it, and set up a new one.
Turn off, re-sync and delete
- Turn off stops sending changes to the app. You are asked to confirm. Accounts already in the app stay as they are.
- Re-sync now sends every assigned person again. Use it after you fixed a problem at the app's end. The app is brought up to date within minutes.
- Delete target removes the configuration. Turn the target off first. You are asked to confirm. The evidence stays in the audit log.
While the app still has accounts that Scrydon provisioned and has not deactivated, you cannot delete the target or the app itself. Deleting either would remove Scrydon's record of those accounts, and they would stay active in the app with nothing left to deactivate them. To delete:
- With provisioning on, remove those people's access to the app (unassign them).
- Wait until Attempts shows a deactivate for each of them, usually within a minute.
- Turn provisioning off, then delete the target or the app.
If the app is gone for good and can no longer be reached, contact support.
When the secret or the connection a target uses is deleted, the target shows Not ready and cannot be turned on. Edit it and choose another one.
What happened, and what differs
The Provisioning tab shows the target's status, its last attempt, and how many open differences it has.
-
Attempts lists every call Scrydon made to the app: when, for whom, what it did (create, update, deactivate or delete), and whether it succeeded, failed or is pending. A failed call is retried by itself. When the app asks Scrydon to slow down (a
Retry-Afteron a 429 or 503), Scrydon waits as long as the app asks, up to 5 minutes, before it tries again. An attempt marked No call made failed before Scrydon could reach the app. Whole target in place of a person means it affected every person at that target. -
Drift lists what Scrydon found when it compared the app's accounts with who should be there:
Finding Meaning Missing in the app An assigned person has no account in the app. Not expected in the app The app has an account for someone who should no longer have one. Attributes differ An account's name, email or teams differ from Scrydon's. Active state differs An account is active when it should be deactivated, or the other way round. Scrydon repairs each finding itself. Turn on Show resolved to see past findings too.
Both lists are paged. To keep a record outside Scrydon, use Export CSV: pick a window of up to 92 days, and Scrydon downloads up to 50,000 attempts from it. Each export is recorded in the audit log.
Troubleshooting
| Message | What to do |
|---|---|
| That URL is refused | Use https, with no user name, password, query or fragment, and a host on your organization's egress allowlist. |
| That secret (or connection) is not one of this organization | It was deleted, or belongs to another organization. Choose another one. |
| This app already has a provisioning target | Edit the existing target instead. |
| Turn provisioning off before deleting the target | Turn it off, then delete it. |
| Accounts are still provisioned downstream (or: This app still has accounts provisioned downstream) | Remove those people's access, wait for each deactivate in Attempts, then delete. See Turn off, re-sync and delete. |
| That window holds too many attempts for one file | Export a shorter window. |
An attempt failed with scim_unauthorized or scim_forbidden | The app refused the token. Put a token the app accepts in the secret. |
An attempt failed with scim_endpoint_refused | The base URL is no longer allowed, often because its host left the egress allowlist. |
An attempt failed with scim_uniqueness | The app has another account with that userName. Remove the duplicate in the app, or give it the person's externalId. |
An attempt failed with scim_rate_limited, scim_target_unavailable or scim_timeout | The app is busy or down. Scrydon retries by itself. |
Microsoft Entra
An Entra target writes your people into your Microsoft Entra tenant through Entra's API-driven inbound provisioning. Scrydon calls it through your organization's Microsoft extension connection, so it never holds a Microsoft token of its own.
Before you set it up, in your Entra tenant:
- Create the provisioning app and its API-driven provisioning job (Enterprise applications → API-driven provisioning to Microsoft Entra ID). Note the app's service principal object id and the job id.
- Grant the app registration behind your Microsoft connection exactly two
Graph application permissions, with admin consent:
SynchronizationData-User.UploadandProvisioningLog.Read.All. - In the job's attribute mappings, map
externalIdto the attribute the job matches on (by defaultemployeeId),userNametouserPrincipalName, andactivetoaccountEnabled. - Start the job.
Try it in a test tenant with a few people first.
What Entra receives. For each person, one SCIM bulk request with
userName (their sign-in name, which becomes their user principal name),
externalId (their fixed identifier), their name, one primary work email,
and active. Deactivating sends active: false: Scrydon never deletes the
Entra user. Teams are not sent to Entra.
Pending is normal. Entra accepts an upload and applies it later, in its next provisioning cycle. Until then the attempt shows pending. Scrydon reads the job's provisioning log and moves the attempt to succeeded or failed when Entra has applied it. A failed one is sent again.
Limits. Entra accepts 40 uploads every 5 seconds and 2,000 every 24 hours for a tenant. Scrydon sends one person per upload and waits when a limit is reached, so a first sync of more than 2,000 people takes more than a day. Every target that uses the same Microsoft connection shares these limits, so use one connection per tenant.
To stop Entra applying uploads at once, stop the provisioning job in the Entra admin center, then turn the target off here.
| An attempt failed with | What to do |
|---|---|
entra_access_denied or entra_reauthorization_required | Grant and consent the two permissions above, or reconnect the Microsoft connection. |
entra_runtime_refused | The Microsoft connection is missing or turned off. Repair it, or edit the target. |
entra_throttled | Entra's limit was reached. Scrydon waits as long as Entra asks, up to 5 minutes, and retries by itself. |
entra_conflict | Another Entra user already has this user principal name or another unique value. Resolve the duplicate in Entra. |
entra_skipped | The job's scoping left the person out. Check the job's scoping filter. |
entra_provisioning_failed | Entra could not apply the change. The attempt shows Entra's own error. Check the job's attribute mappings. |
entra_not_logged | Entra did not report the upload within 4 hours. Check that the job is started and not in quarantine. |