Natoma

• The Snowflake acquisition could close on terms that fold Natoma fully into the data platform, ending it as a standalone product a non-Snowflake enterprise could buy.
• If the deal does not close, Natoma returns to a crowded MCP-gateway field with one named customer and a single 7 million dollar seed, facing platform vendors that can bundle the same tool gating.
• Platform vendors such as Microsoft, Google, and AWS could ship native MCP governance inside agent toolkits enterprises already license, removing the third-party budget line the gateway depends on.
• The core tool-gating capability sits in the densest cell of the agent-security landscape, so competitors and incumbents could match Natoma's policy and discovery features within a release cycle.
• Natoma has published only one named enterprise reference, so a buyer or acquirer relying on customer proof has a single case study plus the Snowflake deal to judge real demand.
• The value rests heavily on the founding identity team, so departures during or after the acquisition could erode the integration and authorization depth that justified the purchase.

Natoma sells to the enterprise security team that wants to deploy AI agents but cannot see or control what those agents reach for. The platform frames the problem as AI agents and MCP connections spreading across a company faster than security can track, including shadow AI that nobody approved, where an agent can touch production systems and expose data it was never meant to see.

Independent reporting corroborates the gap beyond vendor marketing. CIO and InfoWorld quote analysts who describe MCP as the connective tissue for enterprise agents that turns into shadow-AI risk without identity, policy, and auditability, and Snowflake built its acquisition rationale on the same need to govern what agents can do. The buyer pain is described by outside voices, not asserted by the vendor alone.

Natoma states the consequence in concrete terms. The vendor says agents should act with exactly the permissions of the person behind them, which is the control its authorization and discovery are meant to enforce before an agent runs loose inside core systems. [s7, s8, s1]

The Natoma platform centers on a governed gateway for the connections AI agents use to reach company tools. The product page describes identity-aware authorization, attribute-based policy that defines who can use which tools and under what conditions, managed or bring-your-own credentials, data filters, and context-aware enforcement keyed to user, group, device, and AI client.

The platform pairs that control layer with discovery and audit. Natoma describes finding and inventorying MCP connections across the organization, including unsanctioned shadow AI, plus a complete audit trail of tool access and integration with SIEM, EDR, and MDM systems. The platform page cites a verified library of thousands of MCP servers and granular authorization via Cedar, with support for custom servers across cloud, desktop, and self-hosted environments.

The acquisition itself is strategic validation by Snowflake, the pending acquirer, rather than an independent technical evaluation. Snowflake describes the MCP Gateway as unlocking centralized governance and says Natoma's platform is already deployed at some very large enterprises, an acquirer's endorsement of market value. Natoma also holds a granted patent in the category. The US Patent and Trademark Office granted Natoma Labs patent US 12,615,260 B1, titled System for managing non-human identities, in April 2026 on a January 2024 filing that names co-founder Pratyus Patnaik as inventor, and the analyst tracker CB Insights describes the platform as discovering, monitoring, and managing non-human identities such as service accounts, access tokens, and API keys. [s2, s15, s5, s16, s17]

Natoma competes for the governed-MCP layer against both startups and the platforms moving to own it. Other MCP-gateway companies cover the same tool-gating and governance stack for the enterprise buyer, while SaaS vendors and hyperscalers race to make themselves the orchestration and governance layer for enterprise agents, per analysts quoted by CIO. The analyst tracker CB Insights places Natoma in the non-human identity space and lists access-governance vendor ConductorOne among its competitors, which fits the identity-governance heritage the founding team brought from Okta.

Natoma's visible edge was a credible identity team and an early, clean entry into the category. Founders from Okta and other identity and cloud companies gave it standing, and a library of prebuilt MCP servers gave buyers a fast path to value. That head start is what a data-platform acquirer paid for.

The structural risk is who owns the buyer, and the pending Snowflake deal points against independence. A platform that already holds the enterprise's data is the natural owner of the agent's permissions too, and Snowflake acted on exactly that logic by signing a definitive agreement to buy Natoma rather than competing with it. [s7, s5, s9, s16]

Natoma's go-to-market leans on the agentic-AI wave and inbound interest, routing prospects to a demo request and building awareness through MCP content. The published proof now includes one named enterprise: HPE Networking CIO Venky Rangachari describes Natoma as the framework his team uses to inventory, qualify, secure, and manage agents end to end, starting with 500 ChatGPT users connected to Salesforce and Gong.

The buyer-side proof is decisive. Snowflake signed a definitive agreement to acquire the company about fourteen months after launch, and its own announcement says Natoma's platform is already deployed at some very large enterprises, so enterprise validation is arriving both as a reference customer and as an acquisition.

Investor conviction reinforces the picture. Index Ventures and Greylock led the 7 million dollar seed, and the founders' Okta exit gave the round senior pedigree. One named customer story is still thin proof of broad reach, but it moves the record past investor signal alone. [s13, s14, s5]

The founders pair identity-product experience with a prior exit in the same domain. Co-founder and CEO Pratyus Patnaik sold his earlier startup, atSpoke, to Okta in 2021 for 79.3 million dollars and then spent more than two years as a senior director at Okta before starting Natoma. Co-founders Paresh Bhaya, Will Potter, and Zachary Hart came from Okta, Microsoft, Google, and Salesforce.

That background sits directly under the product Natoma built. Authorization, credential governance, and non-human identity are the problems this team worked on at established identity and cloud companies, so the move into agent tool access is continuous with their prior work rather than a reach.

Greylock's Saam Motamedi and Index's Shardul Shah backed the company on that basis, framing non-human identities as a core part of the enterprise identity perimeter. The pedigree is verifiable through a named acquisition and named investors, not just titles. [s6, s10, s11]

Natoma publishes a trust center at natoma.ai/trust, hosted on Vanta, that lists SOC 2 compliance with a downloadable SOC 2 Type II 2025 report behind an access request, GDPR alignment, and CCPA compliance, and it states that penetration testing is performed. No ISO 27001 attestation appears. That collateral matters because the product sits in the path of agent access to sensitive systems, and the platform pairs it with the audit trails, identity-aware policy, credential governance, and SIEM integration a security buyer evaluates.

The pending acquisition reframes the readiness question. Snowflake plans to integrate the gateway into its governed data platform, so the trust story shifts from Natoma's standalone certifications to how Snowflake governs the capability at enterprise scale, a point CIO's analysts flagged as the real test. [s12, s2, s7, s1]

Add analyzed competitors to compare them side by side with Natoma.

Natoma targets the enterprise security and platform team standing up AI agents at scale. The buyer is the organization where agents already connect to on-prem, cloud, and hybrid systems and where security cannot see or control what those agents reach for. HPE Networking is the named instance, with a CIO-led team that needed a single platform to inventory, qualify, secure, and manage agents across chat, coding, and autonomous use.

The product is built for that buyer rather than the small team experimenting with one agent. The platform page lists VPC deployment, on-prem support, SIEM and IAM and EDR and MDM integration, and audit export, which are the controls an enterprise security review demands before agents touch production. Snowflake describes the platform as already deployed at large global enterprises, which corroborates the enterprise focus beyond Natoma's own pages.

The free tier widens the top of the funnel without changing the target. A team can start with 5 MCP servers and 5,000 tool calls a month at no cost, then grow into Pro and Enterprise as governance needs appear. The motion is land in a developer or platform team and expand into a security-governed deployment, which fits a buyer who adopts agents before the controls catch up.

Natoma's claimed advantage is sitting in the tool-call path rather than watching from the side. The platform is a governed MCP gateway and proxy through which AI clients and agents reach enterprise tools, with identity-aware, attribute-based authorization enforced through Cedar policy and OAuth 2.1 authentication. Every Natoma-mediated tool call routes through the gateway, so policy is applied at the moment of access rather than reconstructed afterward from logs.

Discovery and audit pair with the enforcement layer. Natoma finds and inventories MCP connections across the organization, including unsanctioned shadow AI, and the platform page reports detecting on average 225 unmanaged shadow-AI instances per enterprise. The product produces a complete audit trail of tool access and integrates that activity into SIEM, EDR, and MDM systems, which is the visibility a security team needs to bring shadow connections under one policy.

The verifiable footprint is strong for an emerging category, with deployment metrics and an enterprise reference that corroborate real usage. What holds up, though, is the gateway position and the Cedar-based policy engine rather than a model advantage, since the connectors and authorization logic are engineering a funded rival could rebuild from the same MCP standard. No proprietary model or named non-public dataset appears in fetched sources.

Natoma runs an inbound, demo-led motion riding the surge of enterprise MCP adoption. The homepage routes prospects to a demo request and frames the product around letting AI safely reach everything a company knows, and the free tier gives a developer or platform team a no-cost entry before a security-governed expansion. The founder's Okta pedigree and Index and Greylock backing gave the early round senior credibility.

Named demand is thin but real. HPE Networking is on record through its CIO describing a deployment that began with 500 ChatGPT users connected to Salesforce and Gong and grew into governing chat, coding, and autonomous agents. Snowflake's announcement states the platform is already deployed at large global enterprises, though it names no additional logos, so the public roster rests on one detailed customer story plus the acquirer's unquantified claim.

The decisive go-to-market signal is the acquisition itself. Snowflake signed a definitive agreement to buy the company within about two years of its 2024 founding, which validates enterprise demand for the layer more strongly than investor conviction alone. What it does not settle is whether Natoma would have converted that demand into a broad independent customer base, since a platform vendor moved to absorb the layer before that proof accumulated.

Natoma publishes pricing, which is rare for an enterprise security product and gives buyers a budget anchor. The Free plan covers 5 MCP servers, 5 users, and 5,000 tool calls a month at no cost, Pro raises those limits and adds SSO, access policies, and a 99.99% uptime commitment, and Enterprise adds unlimited tool calls, on-prem deployment, bring-your-own vault, DLP filters, and a 99.999% uptime commitment. Pro and Enterprise prices are quoted through sales rather than listed.

The metering reveals what Natoma believes buyers pay for. The company charges by the number of MCP servers, users, and tool calls, and states plainly that it does not charge per agent, so a customer can deploy any number of agents within plan limits without per-agent fees. That choice prices the product like infrastructure billed on throughput rather than a tool billed on seats, and it positions the meter on connection volume, the unit that grows as agents proliferate.

The published unit also signals the durability case. By tying cost to tool calls on the metered Free and Pro plans, Natoma earns more as a customer pushes more agent activity through the path it controls, which aligns revenue with the inline position. Enterprise breaks that link by including unlimited tool calls, so high-volume deployments are priced on a custom basis rather than by throughput. The tradeoff is that throughput-based pricing exposes the customer to cost as usage grows, which the activity monitoring and budget alerts in higher tiers are meant to manage.

Natoma is a platform offered in managed cloud, customer VPC, and on-prem deployment modes. The platform supports hosted and desktop MCP servers, runs in a customer VPC, and offers on-prem deployment in the Enterprise tier, so a buyer can place the gateway where its security posture requires. The product is software rather than a managed-expertise service, and the platform page says a customer can run Natoma's MCP gateway infrastructure inside its own environment, so the buyer consumes the platform and sets the policy without an analyst-staffed accountability layer.

Operational commitments are unusually concrete for the category. The Pro plan guarantees 99.99% uptime and the Enterprise plan 99.999%, the platform produces audit logs and audit export, and it integrates with SIEM, IAM, EDR, MDM, and OpenTelemetry tooling. Those published SLAs and the integration surface are the operational signals an enterprise change-management review looks for before an access-path component goes into production.

The inline role raises the operational stakes. Because every Natoma-mediated agent tool call routes through the gateway, Natoma becomes a dependency in the agent's access path, so its availability and latency directly affect whether agents can act. The five-nines Enterprise SLA is the company's answer to that exposure, and the bring-your-own-vault and DLP options let a customer keep sensitive credential and data handling under its own controls.

Natoma runs a Vanta-hosted trust center that lists a completed SOC 2 Type II 2025 report, GDPR alignment, and CCPA compliance, and states that penetration testing is performed. The pricing page repeats the SOC 2, GDPR, CCPA, and US Data Privacy claims as procurement assurance. The SOC 2 Type II report is available through an access request rather than open download, which is the standard posture for an enterprise security vendor.

That collateral matters because the product sits directly in the path of agent access to sensitive systems. A gateway that authorizes and proxies tool calls handles credentials, policy, and a record of every action, so a security buyer evaluates the attestations alongside the audit trail, identity-aware policy, credential governance, and SIEM integration the platform provides. The published SLAs and the bring-your-own-vault option add to the formal-review surface.

The attestations are enterprise-grade but table-stakes rather than a moat. SOC 2 Type II and the privacy alignments ease procurement without blocking a substitute, and no compliance regime mandates a governed MCP gateway as a product class. The pending Snowflake acquisition shifts the longer-term trust question from Natoma's standalone certifications to how a data-platform owner governs the same capability at its scale.

Natoma positions itself as the connectivity and governance layer between AI clients and enterprise tools rather than a point control. It standardizes how agents connect to tools through a single governed gateway, maintains an enterprise MCP registry and artifactory, supports custom and self-hosted MCP servers, and works across Claude Code, ChatGPT, Cortex, and custom AI clients. The platform claim rests on owning the access path that agents traverse to reach company systems.

That position is what Snowflake signed a definitive agreement to acquire, a deal still subject to customary closing conditions. Snowflake's stated rationale is to extend its governance perimeter from data assets to AI actions and interactions, connecting Snowflake Intelligence and Cortex Code to enterprise systems through Natoma's verified MCP library and centralizing governance through the gateway. The acquirer wanted the path, not a side monitor, because the path is where data access and AI action meet.

The exposure is that platform owners are the natural owners of this layer. A vendor that already holds the enterprise's data and identity is positioned to fold MCP tool gating into a suite the buyer already licenses, which is the logic Snowflake acted on by agreeing to buy Natoma rather than compete with it. The same logic leaves an independent gateway exposed to hyperscalers such as Microsoft, AWS, and Google, which are consolidating their agentic development toolkits with similar tools and functions as vendors race to own the governance layer.

Natoma's credibility comes from a founder with a prior startup exit and years inside an identity vendor. Founder Pratyus Patnaik previously built and sold his earlier startup, atSpoke, to Okta in 2021 for $79.3 million, then spent more than two years as a senior director at Okta before leaving to establish Natoma. That sale is a verifiable prior exit rather than an unverified background claim, though atSpoke was a workplace operations platform, so the identity-domain grounding comes from his Okta years rather than from what he built before.

Natoma built its governed-gateway platform with a small team, with DataBreachToday reporting 27 total employees and The Register reporting that the pending deal would bring 20 employees to Snowflake. Patnaik has a prior enterprise software exit to Okta, and Snowflake credits Natoma's team with expertise in MCPs, gateway infrastructure, identity governance, and privileged access management, so the move into agent tool access builds on that base.

The backer and acquirer signals reinforce the team read. Index Partners and Greylock led the seed, and Snowflake's own announcement credits the team with deep expertise in MCPs, gateway infrastructure, identity governance, and privileged access management. The verifiable strength is the founder's domain track record and the confidence it drew, with depth below the principal the open question fetched sources do not resolve.