Nexos.ai vs Portkey (2026): AI Gateways Compared — and Where Kosmoy Fits

Three shapes on the same ten axes. nexos.ai (orange) holds a distinctive profile for a gateway vendor: 8 on Gateway & Policy Control, 7 on Guardrails and a workspace-driven 5 on Agent Building, but it drops to 4 on Deployment Sovereignty — vendor-hosted only — and 6 on Observability with short log retention. Portkey (violet) peaks at 9 on Gateway, Observability and Sovereignty, with 8 on Guardrails — the traffic-depth signature. Both cluster at 5 on the inventory and compliance axes. Kosmoy (blue) trades a little raw gateway breadth for reach across inventory, compliance and agent containment. Read it as area: the two gateways compete spoke by spoke; the suite covers the web.

One governed surface for the whole company. The Workspace gives employees chat across 200+ models, no-code AI Agents and shared Projects on the same governed platform as the developer gateway (features). Portkey governs traffic for builders; it has no employee workspace and no agent builder of its own — it governs agents built elsewhere.

EU residency, zero data retention and an ISO 42001 certificate. An EU-hosted gateway with zero-data-retention processing and SOC 2 Type II, ISO 27001 and ISO 42001 certifications (EU gateway). Portkey's documented posture is SOC 2, ISO 27001 and HIPAA — without ISO 42001 or an EU-residency pitch.

Adoption without a platform team. Per-seat workspace tiers plus a pay-as-you-go gateway (pricing) roll out like a SaaS app to the whole company; Portkey's free-to-Pro-to-enterprise ladder is priced for the platform and engineering teams that operate it, not per employee seat.

First-party PII guardrails, on by default. Input/output filtering and real-time PII blocking/redaction (emails, cards, IBANs, location) ship as part of the platform, scoped per team, tool and model (AI Guardrails) — no engine selection required, where Portkey pairs its 20+ built-in checks with partner engines (Zscaler, Akto, Lasso) for advanced detection.

Provider and model breadth. One API to 250+ LLMs and 1,600+ models across 45+ providers (gateway repo) against nexos.ai's 200+ models — and the gateway core is MIT open source, adoptable without procurement.

Observability depth. Full request logging, 21+ metrics, per-key budgets, OpenTelemetry and data-lake export (observability docs) — a 9 against nexos.ai's 6, where request logs are kept 7 days, metrics 90 days, and no OTel export is documented (changelog).

A deployment ladder down to air-gapped. SaaS, hybrid with the data plane in your VPC, open-source self-host, or a documented fully air-gapped enterprise tier (self-hosting docs). nexos.ai offers no customer-managed deployment as of August 16, 2026 — its “private hosting” is isolated tenancy inside nexos's own EU infrastructure.

MCP and agent gateway maturity. An MCP gateway with per-user tool provisioning and, since April 2026, an agent gateway with skill-level access control (MCP gateway docs). nexos.ai's MCP support is a curated set of managed integrations (Atlassian, GitHub, GitLab) — no general-purpose MCP gateway with tool-level policy for arbitrary servers is documented.

nexos.ai answers “how do we give employees and developers governed AI access, in Europe?” Portkey answers “how do we route, observe and guard model traffic, wherever it runs?” Neither answers “what AI is running across the organization, can we prove it is compliant, and what happens when an agent misbehaves?” Both inventory only what flows through their own platform, neither documents EU AI Act evidence tooling — nexos.ai's ISO 42001 certificate covers its own management system, not your evidence — and neither documents agent sandboxing or a kill switch as of August 16, 2026. When those are the ask, the category changes — from gateway to control plane, the frontier mapped on the AI capability map.

Kosmoy runs single-tenant in your Kubernetes and includes the gateway both vendors sell — one OpenAI-compatible policy point with guardrails, RBAC, budgets and logging — but wraps it in the layers a gateway leaves out: a risk-tiered inventory of every model, MCP server and agent across the estate; EU AI Act, ISO 42001 (aligned) and NIST AI RMF evidence built from registry state plus gateway logs; and kernel-enforced Action Capsule containment for agents that act. The direct head-to-head lives at Kosmoy vs Portkey.

So the honest framing is not “Kosmoy beats nexos.ai and Portkey at being a gateway” — nexos.ai is a stronger employee front door and Portkey routes to more models with deeper observability. It is that a gateway covers two or three spokes of the ten-axis map. If the requirement is the whole web — inventory, gateway, compliance and containment in one self-hosted platform — that is a suite decision, not a gateway decision.

For a team whose problem genuinely is governed AI access, pick on the axis that decides it: nexos.ai if EU residency and company-wide adoption matter most and a vendor-run platform is acceptable; Portkey if traffic depth, observability export or self-hosting decide it. Both expose OpenAI-compatible endpoints, so application traffic moves with a base-URL change — the sticky parts are the employee workspace habit on one side and gateway configs, logs and prompt templates on the other.

For an enterprise that has to prove control over all of its AI — not just route it — the choice is not between these two but between a point tool and a suite. Kosmoy can coexist with either: some teams keep nexos.ai as the employee chat surface, or Portkey as a developer gateway, while Kosmoy holds the estate-wide inventory, compliance evidence and containment for what reaches production.