Microsoft 365 security for MSPs

Find what's wrong in every client tenant, fix it with a runbook, and prove it's fixed.

A grounded assessment across your whole book, paste-ready runbooks for the fixes, and agents that prepare each fix and verify it held — on the Microsoft 365 your clients already own.

TrueRock.AI is a multi-tenant Microsoft 365 security platform for Managed Service Providers (MSPs), built by TrueRock AI LLC in Redmond, Washington.

Every Microsoft 365 planMulti-tenantGrounded & cited

Read the narration
  1. TrueRock finds what's wrong in a client's tenant, fixes it with a runbook, and proves it's fixed.
  2. Here is one Northwind client, Prism Healthcare. 250 users, assessed end to end.
  3. Conditional Access has 2 gaps, and 22 of 84 devices are out of compliance.
  4. First, who can already reach the data. 7 SharePoint sites are open to everyone.
  5. Only 12.9 percent of files carry a label. And a label is what the AI reads before it answers.
  6. 19 agents run in this tenant. 10 hold tenant-wide write access, and 3 have no enabled owner.
  7. At the top, EHR Files Sync. Its only owner is disabled, and it can write every file in the tenant.
  8. TrueRock writes the runbook to fix it. A dry run first, and then a technician approves it.
  9. In this example, an agent reads the tenant again, and the problem is no longer there. That is the proof.
  10. The risk you find becomes the fix you prove. For Prism Healthcare, and for every client in the book.
What it is

One console for your whole client book: assess, fix, prove.

TrueRock OutCrop assesses every client tenant you manage, hands your team runbooks for the fixes, and puts agents to work preparing those fixes and checking that they held.

Why “OutCrop”

In geology, an outcrop is where the bedrock breaks the surface — the one place you can see what the ground is really made of. TrueRock OutCrop does that for every client you manage. It brings what's hidden inside each Microsoft 365 tenant to the surface — overshared files, shadow AI, over-privileged agents, missing Conditional Access, configuration drift — ranked across your whole book, with the evidence behind every finding and the runbook to fix it.

OutCrop · msp.truerock.ai

The security console

// every client tenant, assessed and ranked

  • Every client ranked worst-first, one screen
  • Identity, devices, data and agents — each finding traced to its evidence
  • Paste-ready runbooks for the fixes, dry run first
  • Compliance mapped to the frameworks, cited
OutCrop · Quarry agents

The agents and the Review Desk

// fixes prepared, approved, verified

  • Agents prepare each fix with its runbook; Obsidian, the investigator, remembers which fixes held
  • A technician approves it on the Review Desk
  • The tenant is read again to verify the fix held
  • A security brief shows the client what was fixed

Optional: a governed AI workspace for clients who want one. TrueRock Managed AI →

Assess every client. Fix it with a runbook. Prove the fix held — and show the client. How pricing works →
Why it matters

Find it. Fix it. Prove it.

What it means for you — and for the clients you protect.

For the MSP
A recurring security service, with proof of work.

Assess every client on the Microsoft 365 they already own, work the fixes from one Review Desk, and show each client what was fixed and verified — month after month, across your whole book. The work your team does stops being invisible.

For clients
See where you stand — and what was fixed.

A plain security brief: where your tenant stands, the top risks, and what your partner fixed — each fix checked against your own tenant, not taken on trust. Every finding is cited to its evidence, so you can hand the record to your board or your auditor.

See it work

Every capability, with the demo that proves it.

Pick a product; watch the film — every demo cited to what it shows.

Start here: the 5-minute tour
One client’s dashboard, opened from your whole book; a fix with its runbook; the AI insight checked statement by statement; and the Quarry crew that prepares the work.
Findings search
From a question to a safe, ready-to-run fix in seconds.
Your technicians’ AI
Ask in Claude. TrueRock answers from your clients’ real data, with runbooks and PowerShell filled in.
Why TrueRock

Insights down to bedrock, not a checkbox-tick.

A one-time readiness check is a snapshot — stale the day it ships. TrueRock is built to reason deeper, prove every finding, and keep measuring across your whole book.

01

Depth, not a snapshot

Reasons over a client's real environment — the connections and exposures a flat checklist never sees.

02

Grounded, not guessed

Every finding traces to its evidence. "The model said so" doesn't survive a client review. Provenance does.

03

Continuous, not one-time

The same intelligence never stops — the first assessment through every day you manage the tenant.

04

Built in, not bolted on

The Overview, the runbooks, the agents and Ask OutCrop all read the same facts — not a chatbot bolted onto a legacy dashboard. Multi-tenant from day one, so it spans your whole book with clean, GDAP-secured isolation.

On the Microsoft 365 they already own.

Help every client get the most out of the Microsoft 365 they own. On every plan from Business Basic up, TrueRock assesses each client in layers — identity, devices, threats, data and sharing, email, AI agents, licensing, your baseline and compliance — with curated runbooks for the fixes and agents that check they held, across your whole book. It works alongside the Microsoft 365, Copilot and E5 your clients already have, and reads their richer signals where they exist.

✓ Business Basic
✓ Business Standard
✓ Business Premium
✓ E3 — and richer where signals exist
✓ E5, Copilot and Agent 365: their signals read in

What TrueRock adds on top of the Microsoft 365 your clients already own — capability by capability.

See the capabilities in detail 21 capabilities

The security and compliance coverage TrueRock adds on top of the Microsoft 365 your clients already own. Your Microsoft 365 plan = what the client’s plan provides today · Plus TrueRock = what TrueRock adds on every plan, across every client. Managed AI is an optional add-on, listed without a comparison.

OutCrop 19 The security and compliance coverage TrueRock adds on top of the Microsoft 365 each client already owns — assessed across your whole book, every finding cited, with runbooks for the fixes.
Capability Your Microsoft 365 planthe Microsoft SKU, natively Plus TrueRockon any SKU · cross-tenant
A layered Overview of every client Each client tenant opens on one Overview: Identity, Devices, Threats, Data & sharing, Email, AI agents, Licensing, Baseline and Compliance, each a tile with its findings and a Visual view. Every plan gives admin centers and a Secure Score for one tenant at a time, each in its own portal. Basic → E5One Overview per client in the same nine layers, worst first, each tile opening on the findings behind it, and honest about what it could not read. Each layer’s AI insight runs under embedded observability: objective judges score every statement against the assessment, and the insight says how many it verified. A portfolio view puts every client side by side.
Portfolio Insights across the book Four views of your clients at once: where each client sits on two measures you pick, a grid of controls by client, what changed in each client’s Secure Score, and every client that shares the same gap. The admin centers and Secure Score report each tenant in its own portal. Basic → E5Each client drawn from its latest saved assessment, and every dot, cell and row opens that client in the dashboard: its Overview, the finding or check behind a control, the layer that holds it, its changes. A measure that was not read is listed with its reason instead of plotted.
Finding drawers with paste-ready runbooks Open any finding, or the user, device or app behind it, and read what it means, the evidence, and the curated runbook that fixes it. A finding with no curated runbook yet says so. Admin centers show a setting and Microsoft’s guidance for it; the technician writes and checks the script. Basic → E5Paste-ready PowerShell filled in for that client, with a check-only copy that changes nothing, separate from the apply copy wherever the script has both, and the permissions, steps and rollback the runbook records.
Findings search Every finding for a client in one list, searchable by finding, object, layer or fix. Findings are spread across the Defender, Entra, Intune and Purview portals. Basic → E5One searchable list per client; each result opens in the same drawer as the Overview.
Identity and Conditional Access Users, admins and the sign-in policies that protect them, and the users a policy misses. Conditional Access from Business Premium up (Entra ID P1); Basic and Standard rely on security defaults. E5 adds risk-based policies (Entra ID P2). Basic → E5Coverage gaps per client, license-aware: the failing policies and the users left out, and on the plans that cannot enforce Conditional Access, the gap said plainly.
Devices Managed devices against compliance, encryption and antivirus health. Intune from Business Premium up; E5 adds Defender for Endpoint. Each reports one tenant. Business Premium → E5Device compliance for every client in the Devices layer, each failing device opening on its finding, and on its runbook where one exists.
Threats: the CVEs that apply The CVEs that match each client’s products, with known-exploited ones called out, and the Defender alerts the tenant records. E5 adds Defender vulnerability management for one tenant; the business plans have no CVE view. Basic → E5The applicable CVEs for every client, with the action that clears each and the devices it applies to, and Defender’s signals read in where a client has them and grants access.
Data & sharing Sensitive data, external sharing, and who and what can reach it across SharePoint and OneDrive. Sharing reports for one tenant on every plan; E5 adds Purview auto-labeling and data-access governance. Basic → E5Sensitive data and its exposure for every client, the people and apps that can reach it, and the label or sharing change to make, as a runbook.
Email authentication SPF, DKIM and DMARC for each client’s custom domains. Every plan supports SPF, DKIM and DMARC; checking them across clients is manual. Basic → E5Each client’s domains checked in the Email layer, with the record to fix and its runbook.
AI agents and their credentials The apps and agents connected to each tenant, including the ones clients build themselves, with their permissions, owners and the static credentials they still use. Entra lists app registrations and consents for one tenant. E5 and Agent 365 add richer agent signals. Basic → E5Every app and agent ranked by what it can reach, findings with an agent runbook opening on the PowerShell that revokes exactly the grants flagged, and Agent 365 signals read in where a client has them and grants access.
Licensing Seats bought against seats used, and the license changes that reclaim waste or close a security gap. A license list per tenant in the admin center. Basic → E5Per-client advice: reclaim unused seats, secure the users a plan leaves exposed, and recommend an upgrade only where a gap needs it.
Your security baseline Every client measured against your baseline: a CIS Level 1 starter you can edit, with per-client exemptions that record their reason. Secure Score and CIS guidance per tenant; a baseline across clients is kept by hand. Basic → E5One baseline for your whole book. Leave controls out or add CIS recommendations, exempt one client from one control with a reason, and the Baseline layer shows each exemption instead of counting it.
Compliance and alignment review Framework controls mapped to each client’s real configuration (CIS, NIST, ISO 27001, HIPAA, PCI DSS, SOC 2, Essential Eight and more), each traced to its evidence. Compliance Manager scores controls for one tenant, from Business Premium up. Basic → E5Each requirement marked aligned, misaligned, not measured or not applicable for every client, with the findings behind it. Controls that need a person to attest are named as such, not scored.
Quarry agents and the Review Desk Agents read each client’s findings and prepare one card per change that has a curated runbook: the case, the evidence, and the runbook with its dry run and rollback. Not part of Microsoft 365 today: no plan prepares fixes across an MSP’s clients for a technician to approve. Basic → E5Approve or reject each card. Hand-off checks the client for drift, then sends the change to your tool or ticket; Plumb reads the client again to verify the fix and checks later that it held, and Obsidian, the investigator, can say in the next case whether an earlier fix held. Cards retire when a newer assessment no longer finds the problem.Not in Microsoft 365 plans today
Client week The week’s ranked fixes for one client, then what changed in the last 30 days and whether it keeps happening. Audit logs per tenant; no weekly plan. Basic → E5Each fix with its runbook’s time and the standards it helps, opening its card in the same drawer.
Security brief for the client A brief for each client: where it stands, what you fixed, its top risks, the decisions you need from it and what you fix next. Not part of Microsoft 365 today: no plan writes a client-facing record of an MSP’s work. Basic → E5An Executive view for the client’s leadership and a Technical view with control ids and runbook links. Every figure comes from a recorded card or a saved assessment. Copy it as text or print it.Not in Microsoft 365 plans today
Ask OutCrop Ask about any client in plain English and get an answer from its assessment, with citations to the records behind it. The admin centers answer for one tenant at a time. Basic → E5Answers from the same facts the Overview shows, with citations, under embedded observability: objective judges score each answer against the client’s records before you see it, and an answer the judges could not score is marked not verified. A question blocked for sensitive data says why and what to do. When an answer is about a finding with a runbook, the runbook opens in the same drawer.
Onboard the whole book in one consent One GDAP admin consent registers every client you manage; a plan quota caps the scan before it starts. GDAP is Microsoft’s own and you already have it. TrueRock adds nothing to the mechanism. Basic → E5One consent covers your book, not one tenant at a time. Pick a pilot client, run it, and see what was read and what was not.
Findings into your PSA, as tickets Turn findings into tickets in ConnectWise, Autotask or HaloPSA, routed by rules you set. Not part of Microsoft 365 today: no plan writes a ticket into an MSP’s PSA. Basic → E5Routing rules (severity and control) pick the board and queue. Connecting takes your own PSA credentials.Not in Microsoft 365 plans today
Add-on: Managed AI 2 Optional add-on, switched on per client: a governed AI workspace for the client’s people, every AI call made through TrueRock metered to that client.
Capability Optional add-onswitched on per client
A governed AI workspace For clients who want one: their people ask questions of their own documents, every answer cited to the file, sensitive data redacted first. Optional add-onOff until you switch it on for a client. Priced per client, never per seat.
Connect the AI your clients already usesoon Coming: your clients connect the AI they already use, signed in with Microsoft or with access tokens you issue. Optional add-onEach client sees only its own data, under your policy, within the credits you allocate.

See how TrueRock is priced →

Start with one client

Show every client where they stand — and what you fixed and verified this month.

It starts with a single tenant. Connect one client and TrueRock returns a fully cited assessment — identity, devices, exposed data, agents and compliance gaps — every finding traced to where it lives, with runbooks for the fixes, and honest about what it could not read. That first read is the conversation that becomes a recurring security service.

Rather ask a question first? .