Foundations·12 Nov 2026·6 min read

PAM for MSPs: One Team Running Privileged Access in Every Customer Tenant

A managed service provider runs on one engineering team and dozens of customer environments, and the usual glue (shared logins, a spreadsheet of VPNs, orphaned personal accounts) is exactly how third-party breaches happen. Monopam makes the provider-to-tenant relationship a governed object: group-based grants with approval, short-lived signed assertions, and an audit trail the customer owns.

PAM for MSPs: One Team Running Privileged Access in Every Customer Tenant

The economics of a managed service provider are simple: one engineering team, many customer environments. The security consequence is just as simple, and much less discussed: every credential that team holds is a key into somebody else's infrastructure. The Verizon DBIR 2025 puts third-party involvement at roughly 30% of breaches, a figure that doubled year over year, and regimes like NIS2 now ask pointed questions about exactly this kind of supply-chain access. The MSP is not a bystander in that statistic; the MSP is the third party.

The workarounds everyone recognizes

Ask how a typical provider actually reaches customer systems and you get three answers, usually all at once. A shared admin login inside each customer environment, because creating real accounts for every engineer was too much friction. A spreadsheet (or a wiki page, or a password manager vault labeled "customers") holding VPN profiles and passwords. And, at the more diligent shops, a personal account per engineer per customer: better on paper, except nobody deprovisions them when an engineer leaves, so the access outlives the employment.

Each workaround answers a real operational need. None of them can answer the questions a customer, an auditor, or an incident responder will ask: who can reach this environment right now, who approved that, and what did they do while they were in?

Provider access as a first-class object

Monopam's answer is to stop treating provider access as a pile of borrowed credentials and make the provider-to-tenant relationship itself a governed object. It supports this in both hosting models: customer tenants running on the provider's own installation, and customers who run Monopam themselves.

Model A: tenants on your installation

When the provider hosts the tenants, engineers are granted into a customer tenant by group, with an approval step, and once inside they work under built-in roles: Tenant Admin, Operator, or Auditor. Auditor is read-only, and the restriction is enforced server-side, not hidden in the UI. A central provider-access view shows, across the whole installation, who holds access to which tenants; the spreadsheet question gets a live answer.

The sign-in itself is deliberately unglamorous: a signed, single-use assertion valid for 60 seconds carries the engineer into the tenant, and the grant is re-checked on every request, so revocation takes effect immediately rather than at the next login. Operationally, each customer tenant runs on its own database with pinned application servers, which keeps one tenant's load and data handling separate from the next.

Shared infrastructure accounts get the same discipline. A rotated vault account on the provider's main tenant is referenced by customer tenants, never copied into them: when the password rotates centrally, the tenant simply reads the new value at its next request. There is no second copy to drift stale, and no rotation job fanning secrets out to places that then need their own protection.

And the part the customer cares about most: sessions are recorded in the tenant where they happen. The customer sees everything the provider's engineers did, in their own audit trail, without asking.

Model B: the customer runs Monopam itself

Some customers will never put their privileged access on someone else's installation, and they should not have to. In the remote tenant link, the customer keeps its own database, its own Redis, and its own vault key. The only connection is a WebSocket the customer opens outward; there is no inbound port on the customer side to firewall, scan, or explain to a security review.

Enrollment uses a one-time code, after which Ed25519 and X25519 keys are generated on the tenant and never leave it. Every message across the link is signed, timestamped, and single-use. Most importantly, the customer approves each provider user itself: the provider proposes, the customer decides. For day-to-day work there is also an opt-in entry on the tenant's own sign-in page: "sign in as service provider" routes the provider's engineer through the provider and lands them back in the tenant, authenticated and scoped.

One team, every tenant, nothing borrowed

The MSP model does not have to mean shared logins and tribal-knowledge spreadsheets. With the provider relationship modeled explicitly (grants, approvals, roles, short-lived assertions, and an audit trail the customer owns), one team really can run privileged access for every customer, and prove it. The full picture is on the MSP solution page; for the inverse direction, where you are the one letting vendors in, see vendor access.

Tagsmsppammulti-tenantprivileged-accessmonopam

Ready to start managing
identities the right way?

Spin up a fully-loaded trial tenant in under five minutes. No credit card. No sales gate.