Ivanti Opens Neurons Data to AI Clients via MCP, Governed by Existing Permissions
Ivanti's new MCP server exposes Neurons endpoint and exposure data to Claude, Copilot and Neurons agents under existing OAuth and role-based permissions. Ivanti reports 99% internal Copilot adoption in 90 days and a 483% ROI on automated endpoint management.

Why it matters
- Ivanti announced its MCP connection for Ivanti Neurons on November 18, 2026
- 99% of Ivanti employees actively used Copilot within 90 days of deployment
- Ivanti reports $11.2 million in benefits and 483% ROI over three years from its 10,000-endpoint Customer Zero deployment
- Engaged GitHub Copilot users merge 5.4 times more pull requests 1.9 times faster than peers
- AI access is governed by existing OAuth, role-based access controls and policy enforcement
Ivanti announced on November 18, 2026 a Model Context Protocol (MCP) connection that gives supported AI clients governed access to operational data in Ivanti Neurons, its system of record for endpoint, service and exposure information. The move places Ivanti among the enterprise vendors wiring their data platforms into the MCP standard, an open protocol created by Anthropic for connecting AI systems with data sources, tools, resources and prompts.
The stakes are straightforward for security and IT operations teams. Endpoint, patch and exposure data is among the most sensitive information an enterprise holds, and AI clients that query it directly create a new access path. Ivanti's answer is to make the protocol inherit the permissions that already exist rather than create new ones.
What does the MCP connection actually provide?
The connection is intended to ground AI recommendations, investigations and actions in device, digital-experience, patch, exposure and software intelligence. Ivanti's MCP documentation lists GitHub Copilot, Anthropic Claude and Neurons agents as supported systems, while support for additional clients and custom agents can vary by release.
The MCP server makes authorized Neurons data, tools, resources and prompts available according to configured permissions. OAuth, role-based access controls and policy enforcement continue to govern interactions. Existing platform controls maintain auditability and organizational oversight. Configured permissions therefore set the boundary of each AI-assisted workflow.
Two design decisions define the architecture. Neurons remains the system of record, and MCP functions as a protocol rather than a data store. Ivanti also states that it does not use customer data to train models.
Documented use cases listed by Ivanti include:
- Finding failed patches and overdue scans
- Grouping devices by likely cause of an issue
- Identifying critical exposures for investigation
- Surfacing abnormal endpoint signals, recent changes and fleet patterns
- Summarizing environment health, risk trends and the factors driving change
How did Ivanti test this internally?
The product work follows internal deployments of Microsoft Copilot across the business and GitHub Copilot in engineering. Ivanti reports reaching 99% active Copilot use across the organization in 90 days. It classifies 58% of employees as Power Users and reports that 64% of developers operate in the highest stage of its GitHub Copilot maturity model, where they use multi-agent workflows.
The engineering figures stand out even against the high adoption number. Ivanti reports that engaged GitHub Copilot users merge 5.4 times more pull requests and merge them 1.9 times faster than their peers.
The internal program combines employee training and governance with department-level think tanks and an AI center of excellence intended to turn experimentation into repeatable practices. That structure, rather than tool availability alone, is what Ivanti presents as the driver of its adoption figures.
What new assistive capabilities ship alongside it?
A companion announcement distinguishes native AI capabilities inside Ivanti products from MCP access for supported external systems. The new assistive capabilities include:
- Device-query building
- PowerShell stages
- AI insights and recommendations
- AI reporting
- Patch insights
Administrators retain control over whether and how recommended actions proceed. That control point matters for endpoint management, where an AI-recommended patch or configuration change can affect thousands of machines at once.
What does the Customer Zero deployment report?
Ivanti validates Autonomous Endpoint Management in its own production environment under its Customer Zero approach. The deployment covers more than 10,000 internal endpoints for a distributed, predominantly remote workforce.
Ivanti reports the following results from that implementation:
- $11.2 million in total financial benefits over three years
- A 483% return on investment over three years
- 66,000 manual tasks automated
- 55,600 hours of employee productivity gained annually
Those figures describe Ivanti's own implementation. The Autonomous Endpoint Management product page identifies the financial-benefit, ROI and automation results as coming from a specific implementation, with customer results varying by environment, configuration and usage.
For comparison, the productivity figure equals roughly 27 full-time-equivalent years of work returned annually — a scale that explains why endpoint automation draws investment even when ROI claims vary.
Who controls what under the new design?
The MCP design creates a clear division of responsibility. Ivanti's controls determine what an authorized AI system can retrieve from Neurons. The customer's agreement with the AI vendor governs how returned data is handled once it leaves the protocol boundary.
That division has a practical consequence: the choice of AI client and its contractual terms become part of the governance decision, alongside the permissions and data maintained in Neurons. Customers must weigh AI vendor contracts as part of governance since vendors control returned data handling.
The protocol layer does not extend Ivanti's data-handling guarantees to the AI client. A deployment that grants Claude or Copilot access to exposure and patch data inherits whatever retention, training and processing terms apply under the customer's agreement with Anthropic, Microsoft or another vendor.
Why does this matter for the market?
MCP has moved from specification to procurement question in under two years. Vendors of record systems — endpoint management, IT service management, security exposure platforms — now face customer demand to expose their data to AI clients, and each must decide where the governance boundary sits. Ivanti's approach, inheriting OAuth and role-based permissions rather than building a parallel AI permission scheme, reduces one class of configuration risk.
The internal adoption figures serve a second purpose. A vendor claiming 99% active Copilot use in 90 days and a 483% ROI on automated endpoint management is positioning itself as evidence, not just assertion, for the products it sells. The caveat that these results come from one specific implementation, with customer results varying by environment, configuration and usage, remains part of the record.
The open question for buyers is the one Ivanti itself flags: the MCP server governs retrieval, not downstream handling. Enterprises that connect AI clients to endpoint and exposure data will need to audit their AI vendor contracts with the same rigor they apply to the permissions in Neurons.
Source: unite.ai
More from Rebecca Stone
Show full bio
Correspondent covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.
754 articles
Related articles
- Adobe Opens Creative Cloud to Outside AI Agents via Expanded MCP
- Genetec's free MCP connector puts Copilot, Claude and ChatGPT to work on security setup
- SyncHub says plain-English AI queries are replacing dashboards
- Figma Adds 20 MCP Partners While Its CEO Calls the Protocol Flawed
- Adobe Opens Photoshop and Premiere to Rival AI Agents in 2027 Beta