Managed deployments audit logs
Managed deployments audit logs record actions impacting your managed deployments: who created, updated, stopped, started, restarted, or deleted one, and how its status changed afterwards. Use them to answer "who changed this deployment, and when?" and "why did it stop running?".
This page is for Organization admins who review audit logs. To audit executions, schedules, and workers, see Workflows audit logs.
Before you begin
- Your Organization is on an Enterprise plan. See Audit logs.
- You can open the audit logs in the Admin Panel›Administration›Audit Logs ↗, or you have an Admin API key to use the API.
Understand what gets recorded
Every event has the target type DEPLOYMENT.
| Event type | Recorded when | Extra event metadata |
|---|---|---|
workflow.managed_deployment.created | Someone creates a deployment. | None |
workflow.managed_deployment.updated | Someone changes a deployment's settings or resources. | fields, and requested_roles_sync if the requested roles changed |
workflow.managed_deployment.stopped | Someone stops a deployment. | None |
workflow.managed_deployment.started | Someone starts a stopped deployment. | None |
workflow.managed_deployment.restarted | Someone restarts a deployment. | None |
workflow.managed_deployment.deleted | Someone deletes a deployment. | None |
workflow.managed_deployment.credential.provisioned | We create the credential the worker authenticates with. This happens when the deployment is created. | trigger, principal_type, credential_id |
workflow.managed_deployment.credential.revoked | We delete that credential. This happens when the deployment is deleted. | trigger, principal_type, credential_id |
workflow.managed_deployment.status.changed | The deployment moves from one phase to another, for example from building to running. | previous_phase, phase |
Every event except status.changed records an attempt. The outcome event metadata is success or failure, and on failure the error event metadata names what went wrong. A request that fails before it changes anything is still recorded as a failure.
Each deployment runs under its own service account identified in the credentials event. The service account is kept when you stop, start, or restart the deployment, and removed when you delete it. So those three actions record no credential event.
Deployments created before the move to service accounts used an API key instead. Stopping or restarting one of them also revokes and reissues its key, and the credential events record it. For these, principal_type is api_key.
Read a log entry
This entry describes a user creating a deployment:
{
"log_id": 1042,
"created_at": "2026-11-12T09:41:07.512Z",
"actor_type": "HUMAN",
"actor_metadata": {
"user_uuid": "3b0d6a52-9c1e-4f7a-8d2b-6a1f0c9e4b17"
},
"event_type": "workflow.managed_deployment.created",
"event_metadata": {
"outcome": "success"
},
"target_type": "DEPLOYMENT",
"target_metadata": {
"deployment_id": "019e8c85-43cf-7508-bc95-7c1b46d4720c",
"deployment_name": "invoice-worker",
"workspace_id": "5f2a8c41-0b7d-4e63-a9d4-2c8e1f6b3a70"
},
"organization_uuid": "c7a41d2e-6f08-4b95-91a3-5e0d8b2f7c14",
"workspace_uuid": "5f2a8c41-0b7d-4e63-a9d4-2c8e1f6b3a70"
}The Admin Panel shows the same entry in three parts. The Admin Panel can show a name or email in place of an ID.
| Part | What it tells you |
|---|---|
| Actor | Who did it. HUMAN, API_KEY, and SERVICE_ACCOUNT. OTHER is our platform, and appears on status.changed events. |
| Event | What happened, with its outcome and the extra metadata from the table above. |
| Target | The deployment, with deployment_id, deployment_name, and the workspace_id of the Workspace it belongs to. |
deployment_id is missing when a create fails before the deployment gets an ID. The entry then identifies the deployment by deployment_name only.
An updated entry lists the settings that changed in fields, as dotted paths such as spec.revision and resources.replicas. It never contains their values, ensuring secrets stay out of the audit log.
Understand status changes
We record a status.changed entry when the platform reports that a deployment has moved to another phase. A single deploy produces several of them.
| Phase | Meaning |
|---|---|
pending | The deployment is waiting to be picked up. |
queued | The build is waiting for a builder. |
building | Mistral Cloud is building the image from your Dockerfile. |
build_failed | The build failed. |
running | The worker is running. |
failed | The worker failed while running. |
stopped | The deployment is stopped. |
terminating | The deployment is being deleted. |
terminated | The deployment is deleted. |
A successful deploy typically records pending to queued, queued to building, and building to running. A phase can be skipped when it's short, and a failed build ends at build_failed.
These entries have no outcome, because a phase change is an observation, not an operation. For the same reason their actor is OTHER with empty metadata. To find who caused one, look at the created, updated, started, or restarted entry for the same deployment_id immediately before it.
Find what you need
In the Admin Panel, use the Event filter and the Target filter set to DEPLOYMENT. See Search and filter.
To see only what people did, and hide the phase changes, filter on the six operation events: created, updated, stopped, started, restarted, and deleted. With the Admin API, pass one event_type parameter per event:
curl 'https://api.mistral.ai/v1/admin/audit-logs?event_type=workflow.managed_deployment.created&event_type=workflow.managed_deployment.updated&event_type=workflow.managed_deployment.stopped&event_type=workflow.managed_deployment.started&event_type=workflow.managed_deployment.restarted&event_type=workflow.managed_deployment.deleted' \
-H "Authorization: Bearer $MISTRAL_ADMIN_API_KEY"To see only the phase changes, pass event_type=workflow.managed_deployment.status.changed. The Admin API reference lists the other filters, such as before and after.
Troubleshooting
| Problem | Cause | Solution |
|---|---|---|
| No managed deployment events appear. | Your Organization isn't on an Enterprise plan, or managed deployment audit logs aren't enabled for it yet. | Check your plan, then contact your Mistral account team. |
| A deploy produces many entries. | Each phase change is recorded as its own status.changed entry. | Filter on the six operation events, as shown above. |
An entry has outcome set to failure. | The operation was refused or didn't complete. The error event metadata names the cause, for example ServiceNameTaken when the Workspace already has a deployment with that name. | Fix the cause and retry. If you can't tell what it is, contact support with the deployment_id or deployment_name and the created_at time. |
| Stopping or restarting a deployment records no credential event. | The deployment's service account is kept across those actions. | None. Credential events appear on create and delete. |
A status.changed entry has no user. | The platform reports phase changes, not a person. | Look at the previous operation entry for the same deployment_id. |
What isn't recorded
- Reads, such as listing or opening a deployment.
- The values of changed settings.
fieldsrecords which settings changed, not what they changed to. - What your worker does once it runs. Workflow executions and schedules have their own events, listed in the Workflows events reference.
Next steps
- Audit logs reference: every field, filter, and event type.
- Workflows audit logs: the entries for executions, schedules, and workers.
- Managed deployments: deploy and operate a workflow worker.
- Service accounts: the identity each deployment runs under.