Netskope AI Agent Marketplace Adds Shared Capacity Credit Pool

What happened

On October 6, Netskope launched the AgentSkope Marketplace. It is a single console where security and network teams can browse, deploy and manage the company's AI agents, and it pays for all of them from one shared pool of capacity credits.118 The marketplace is open to every Netskope customer now, and the company says more agents are on the way.24

The product has four parts. There is a catalog for finding and sizing agents. Allocation controls spread a committed pool of credits across agents. A calculator estimates how many credits a given workload needs before the customer commits. A dashboard tracks usage and sends alerts to help with future planning.49 Every agent runs under the customer's existing policies and compliance frameworks through Netskope's Zero Trust Engine and the Netskope One platform.1

Chief Product Officer John Martin framed the launch as a fix for procurement. Until now, he said, teams bought agents one product at a time, and each new request started a buying cycle over again before it produced any value. He said the marketplace adds one catalog, one credit pool and one view of consumption on top of AgentSkope's existing governance controls.29

Investors liked it. Netskope shares closed 6.62% higher on the day, and trading volume reached about 5.7 times the daily average.4

The credit model is the real news

Most coverage repeated the press release. The detail that matters is in Netskope's own product documentation. Customers buy monthly capacity credits from a single pool, assign them to specific agents, and can move them around in the console with immediate effect. They are not locked into the split they chose at purchase.5 Netskope measures capacity as an agent's monthly work volume, counted in whatever unit suits that agent, such as case investigations or users.5 The company's FAQ says credits are not tied to any one agent, and new agents draw from the same pool without a new procurement cycle.7

The documentation also splits two roles. An "allocator" assigns credits and watches consumption. A "practitioner" runs an agent day to day.5 That matches how large companies actually buy software: purchasing and governance sit in one place, while the operators are spread across SOC and NOC teams.

In my view, this is the bigger story than the catalog. Netskope is charging for agents by the work they do, not by seat or by SKU. Its marketing makes the point directly, describing agent adoption as "one portfolio decision" rather than a series of separate purchases.7 For a vendor, that has an obvious advantage: once a customer has committed to a credit pool, every new agent Netskope ships can be switched on with no new contract. The marketplace is a distribution channel as much as a convenience.

What is actually in the catalog

The catalog is smaller than the word "marketplace" suggests, and coverage tended to blur this. Netskope's product page lists six agents:

• Insider Threat AISecOps
• DLP AISecOps
• Private Access AIOps
• CCI Insights
• DEM Data Intelligence
• DEM Insights7

They are not all sold the same way. The Insider Threat and DLP agents are tagged as marketplace items. Private Access AIOps is standalone. CCI Insights comes with Netskope's SSE product. The two DEM agents come with higher-tier digital experience management packages.7

That lineup is almost the same as the six agents Netskope announced in May, when it introduced AgentSkope as an architectural layer of Netskope One. At that point five were generally available and the insider-threat agent was in private preview.15 In June the company added an AI Risk AISecOps agent as part of its AI Command Center launch, with general availability expected during the third quarter of 2026.12

Every agent is built, maintained and run by Netskope. Customers do not build or train their own.7 So this is not an open marketplace like an app store or the community registries forming around open-source agent frameworks. It is a curated storefront for one vendor's managed agents. There is one hint of a broader direction: Netskope's survey found that nearly half of respondents want one governance model covering agents built by vendors, built in-house and supplied by managed-service providers.14 The launch materials do not say whether third-party or customer-built agents will ever be listed. An independent analysis noted that the materials also leave out pricing, agent-level performance data and integration details beyond Netskope's own console.5

Where the reporting differs

Most outlets told the same story, but some details don't match.

Launch date. The October press release says AgentSkope launched "in early 2026."18 The dated announcement for AgentSkope was May 5.15 Whether you call May "early" is a matter of taste, but anyone tracking the timeline should use the May date.

Survey numbers. In October, Netskope cited a Cybersecurity Insiders study saying 42% of organizations have alerts, cases or tickets that go entirely uninvestigated for lack of capacity. The same study found that 56% name capacity pressure as the main reason they adopt agents, and 55% rank shared context as the most important platform requirement.7 The May launch used a 40% figure for uninvestigated alerts.15 Some October coverage described the shared-context result as "a majority" and added that 51% see scattered context and signals as an obstacle to resolving issues.1 The numbers point the same way. Still, they come from vendor-sponsored research, and the figures shifting between releases is a reason to treat them as directional.

Valuation. Reported market capitalization ranged from about $7.5 billion to $8.3 billion depending on the outlet and when it measured.48 Financial coverage was cooler than the product coverage. One analysis pointed to a price-to-sales ratio above Netskope's historical median, deeply negative operating margins and net insider selling over the past year.8 Separately, analysts at Morgan Stanley and RBC raised their price targets in September after the company's fiscal second-quarter results.10 The share-price jump says less about this feature than about how much the market rewards any credible agentic-AI revenue story.

How it fits Netskope's agent strategy

The marketplace is the latest step in a year of agent-focused releases. They make more sense read together.

The first step was protocol security. In December 2025, Netskope previewed controls for Model Context Protocol (MCP) traffic, the emerging standard that lets AI agents reach enterprise tools and data. The controls included real-time discovery of MCP servers and clients, Cloud Confidence Index risk scores for MCP servers, an option to block MCP traffic by default, detection of non-human traffic, event logging and data-loss prevention for MCP sessions.1420 In March, those capabilities shipped as the Agentic Broker, part of the Netskope One AI Security suite. The suite also includes AI Guardrails, an AI Gateway and AI Red Teaming, and all four became generally available.1113 A Netskope executive argued at the time that most security tools did not yet recognize MCP traffic at all.19

In May, AgentSkope turned the platform into a host for agents of its own, providing runtime, memory, integration and audit controls.715 June brought the AI Command Center, which discovers AI agents and local models on endpoints and servers.12 In mid-September, Netskope announced controls to stop risky AI agent actions before they run.4 Now the marketplace adds the commercial layer that packages those agents.

The overlap with MCP is worth noting. The CCI Insights agent in the new catalog is marketed as speeding up risk assessments of SaaS apps, cloud apps and MCP servers.7 Netskope is using its own agents to help customers judge whether other agents' tool connections are safe. Securing other people's agents and selling its own both rest on the same policy engine. That is the company's main competitive pitch.19

Why it matters

As companies run more task-specific agents, the hard problem moves away from model access and prompt design and toward metering, spending controls and visibility into what agents are doing.5 Netskope's answer is to make agents work the way cloud capacity already does: buy a committed pool, then let operators spend it wherever demand appears.

This approach has limits. A shared credit pool cuts procurement friction only within one vendor's catalog. A security team that runs agents from several suppliers, or its own agents built on open frameworks, still has several dashboards and contracts. Netskope's own survey shows buyers want governance across all those sources.4 The marketplace delivers that only for agents Netskope builds.

My reading is that this is a well-designed purchasing tool that also locks customers in, and the platform it rests on is the more important asset. If Netskope eventually lists outside or customer-built agents under the same Zero Trust controls, AgentSkope could become a real neutral agent platform. Until it does, it is best seen as a smarter way to sell Netskope's own agents.