ZoomEye returned 286 matches for app="SurrealDB" and 1,479 for title="SurrealDB" on November 4, 2026. The counts are close enough to invite a quick conclusion and different enough to require a careful one.
SurrealDB CVE-2026-121419 affects versions before 3.3.0 and involves MCP BoundSubject identity collision for JWT or Bearer access. A public fingerprint can help find candidate instances. It cannot show whether MCP is enabled, which users authenticate, or whether the token binding is fixed.
| Query | Matches |
|---|---|
app="SurrealDB" |
286 |
title="SurrealDB" |
1,479 |
The app query completed at 2026-11-04T23:44:59Z and the title query at 2026-11-04T23:45:01Z. Both used page 1, page size 1, and sub_type=all. The probes completed successfully.
The title count is higher than the application fingerprint. A title may appear on documentation, a client UI, or a deployment that the fingerprint does not identify. The app count is a more specific product signal, but it still does not identify a version.
The counts support a discovery statement: SurrealDB has an indexed public footprint, and a title search finds more candidate pages than an application query. Neither number is a count of vulnerable databases.
CVE-2026-121419 requires an MCP deployment with JWT or Bearer authentication and a vulnerable session-binding implementation. The external index cannot show those conditions. It cannot see the token issuer, the user identifier, the MCP tool configuration, or whether the server has been upgraded.
Inventory SurrealDB instances through internal configuration management first. Record the version, namespace, database, authentication methods, and MCP listener. Identify every agent or service that connects and the token or credential it uses.
For each session, confirm that the binding key includes a per-user identity. A shared method label such as jwt is not a principal. Verify issuer and audience checks, token expiry, and revocation behavior. Map the tools the MCP server can invoke and the data each user can reach.
Upgrade to 3.3.0 or later. Rotate signing keys and active sessions created before the fix. Test two users with the same authentication method and confirm that one cannot resume the other's session.
Treat the 286 application matches as a candidate list, not a vulnerability list. Resolve ownership and version through authorized systems. The 1,479 title matches may contain documentation and unrelated pages, so do not merge them into the same evidence set.
The public measurement is useful for explaining why MCP identity review matters at internet scale. It is not a substitute for checking the version and token-binding behavior of an actual database.