Portkey alternatives in 2026: pricing, features and where each fits

Ayo Adeniran · Oct 23, 2026 · 5 min read

Portkey earned its following with an open-source AI gateway and a hosted layer for logs, guardrails and prompt management. In 2026 it became part of Palo Alto Networks. Portkey's homepage now reads "Portkey is now PRISMA AIRS AI Gateway, generally available for all enterprises", and Palo Alto announced that general availability on July 16, 2026, describing the product as "the AI control plane for the enterprise."

Whether you are a Portkey customer reviewing your stack or a team choosing a gateway for the first time, it is a good moment to compare. Here is what Portkey offers, the main alternatives, and where each one fits.

What Portkey offers

Portkey's pricing page lists four tiers:

Plan Price Recorded logs Log retention Notable
Open source Free, self-hosted No limit on requests Your own Universal API, routing, retries, fallbacks, load balancing, guardrails
Developer Free 10,000 a month 3 days Gateway, logs and traces, 3 prompt templates
Production $49 a month 100,000 a month, then $9 per 100,000 30 days Alerts, guardrails, unlimited templates, RBAC, simple and semantic caching
Enterprise Custom 10 million+ a month Custom SSO, private cloud and VPC hosting, SOC 2 Type 2, GDPR, HIPAA

Going over the log limit does not block requests; logs past the limit are simply not recorded. Portkey's homepage cites 1,600+ LLMs through one API. You store your provider credentials in Portkey as Integrations, which Portkey describes as "a secure vault for your API keys."

So Portkey is mainly a control and visibility layer over providers you already pay. That shapes which alternative suits you.

The alternatives at a glance

Option Hosting How it charges Strongest at
Portkey Open source, or hosted By recorded logs Observability, guardrails, prompt management
LiteLLM Self-hosted, open source Free; Enterprise by request capacity Running everything in your own network
OpenRouter Hosted 5.5% or 8% fee on credit purchases 600+ models on one balance
Vercel AI Gateway Hosted No markup on tokens; some add-ons metered AI SDK, budgets, provider uptime data
Cloudflare AI Gateway Hosted Core features free; 5% on Unified Billing credits Caching and analytics at the edge
Momor Gateway Hosted 5.5% fee on credit purchases; 5% on your own keys Failover, data rules, workflows

LiteLLM

The closest like-for-like if you used Portkey's open-source gateway. LiteLLM is MIT-licensed, speaks the OpenAI format to 100+ providers, and its open-source proxy includes virtual keys, budgets, rate limits, spend tracking, logging and fallbacks. Enterprise adds SSO, SCIM, audit logs and SLAs, priced by annual request capacity, never per token. You run it, along with the Postgres database virtual keys need.

Fits teams that want the gateway inside their own network.

OpenRouter

A different shape: OpenRouter sells model access itself. One balance reaches 600+ models from 80+ providers at list price, with a 5.5% fee on credit purchases (8% on Business). Bring your own keys and the first $25,000 of list-price inference a month carries no fee, then 5%. Standard includes budgets, admin controls and data policy-based routing.

Fits teams that want breadth and one bill more than a logging layer.

Vercel AI Gateway

List price with no markup on tokens, your own keys at no fee, budgets per team, project, API key or member, and request logs with every routing attempt. Trace drains send OpenTelemetry traces to your own tool on Pro and Enterprise, at $0.05 per 1,000 traces plus $0.50 per GB.

Fits teams on the AI SDK who want cost control and logs without a fee on tokens.

Cloudflare AI Gateway

Analytics, caching and rate limiting at no cost, plus retries and model fallbacks. Log storage on the legacy logs model is 100,000 logs in total on the free plan and 10 million per gateway on paid plans; accounts created after September 24, 2026 use Workers Logs pricing instead.

Fits teams already on Cloudflare who want caching and visibility in front of their providers.

Where Momor fits

Momor Gateway suits Portkey users whose main concern was reliability and control over their own provider keys.

Your keys, in your order. Save several keys per provider, prioritized ahead of Momor's keys or as fallbacks after them. Each prioritized key decides whether Momor's keys may stand in for it, and any key can be limited to certain models, API keys or members. A call on your key costs 5% of its list price.

Failover that holds the conversation together. A provider that fails before answering is stepped over. Reasoning, tool calls and results move to the next provider in its own format, with tool-call ids rewritten so calls stay paired with their results. A broken stream is never stitched to another provider's answer.

Data rules in the router. Zero data retention can be required for one call or held for the whole organization: calls then run only on keys marked as covered by an agreement, and file processing is refused. Call records keep metadata only and are deleted after 90 days.

A record you can query. Every answer carries an X-Momor-Request-Id. Look up any call to see every provider it was tried on, each attempt's tokens and cost, and whether it ran on your key, or sum usage by day, caller, provider and model.

What to weigh: Momor Gateway has no prompt templates, guardrail plugins or semantic caching like Portkey's, and it does not export traces to your own collector today. Its catalog is 12 providers. Momor's own keys serve only the providers Momor resells; the rest run on yours.

How to choose

  • You want Portkey's open-source gateway, self-hosted: LiteLLM.
  • You want model access and one bill: OpenRouter.
  • You want no fee on tokens and per-project budgets: Vercel AI Gateway.
  • You want caching in front of providers, on Cloudflare: Cloudflare AI Gateway.
  • You want your own keys to decide routing, strict data rules and careful failover: Momor Gateway.

Start with the Gateway quickstart, or read how usage and call records work.