Portkey alternative

The Portkey alternative with models included

Portkey gives you a control panel over your own provider keys. Opper is the gateway where the models come with the account: 700+ models, billed through one key, no prompts stored by default, and self-serve EU data residency from your first call.

Get startedView Documentation

Models come with the key

Call 700+ models through one account and one bill, with no OpenAI, Anthropic, or Google keys to collect, rotate, or reconcile. Bring your own keys only if you want to.

No prompts stored by default

Portkey records requests and responses as its core product. Opper meters cost, latency, and tokens, and stores no prompt bodies unless you opt into tracing.

Self-serve EU residency

Run EU-hosted models from your first call, with no hybrid deployment or enterprise contract to set up first.

Portkey fee structure (flat monthly platform fee, metered on recorded logs) per portkey.ai/pricing, accessed August 2026. Opper is pay-as-you-go with no markup on inference, current fees are on the pricing page.

At a glance

Portkey alternatives, compared

Portkey vs Opper feature comparison
FeaturePortkey (Prisma AIRS)Opper
Unified API to leading models
1,600+ vendor-stated
700+
Drop-in OpenAI SDK compatibility
Yes
Yes
Models billed through one account
Bring your own provider keys
BYOK optional
Built-in tracing and observability
Core product
Opt-in via Control Plane
Prompts stored by defaultRequests and responses recordedNo, metadata only
Self-serve EU data residency
Enterprise hybrid or VPC
From the first call
PII and prompt-injection guards
Tier-gated
Yes
Prompt management CMS
Templates, versioning, playground
No
Self-hosted or hybrid deployment
OSS core, VPC data plane
Enterprise plans only
Agent SDK, CLI, and AI Wallet
MCP and A2A gateway
Yes
Platform fee structureMonthly fee, metered on recorded logsPay-as-you-go, % fee on top-ups

Want to compare individual models head-to-head? Run them side by side in the model playground →

Trusted by 50k+ developers and companies serving 10M+ users

Aixia
evroc
GetTested
Instabridge
LexBox
Ping Payments
Steep
Svenska Bostäder

Why teams switch

Why teams move from Portkey to Opper

Portkey built one of the deepest observability gateways on the market, and it remains a strong product. Teams re-evaluating it in 2026 tend to hit four constraints.

The roadmap now runs through Palo Alto Networks

Palo Alto Networks announced the acquisition on April 30, 2026, completed it on May 29, 2026, and made Portkey's technology the foundation of the Prisma AIRS AI Gateway, generally available since July 16, 2026. The standalone product and pricing remain live, and what the long-term roadmap looks like for teams outside the Prisma AIRS platform is not stated. That is good news if you are consolidating on Palo Alto, and a reason to re-evaluate if you chose Portkey as an independent developer tool.

Bring your own keys, and everything that comes with them

Portkey does not resell inference: you plug in your own OpenAI, Anthropic, and Google credentials and keep a billing relationship with every provider you use. Portkey's model catalog manages those credentials well, but the procurement, the invoices, and the rate limits stay yours, and they multiply with every model family you add.

Observability works by recording your prompts

The docs are direct about it: Portkey “records all your multimodal requests and responses”. That depth is the product's core strength, but it means prompt and response bodies sit in a third-party log store by default, with log retention windows of 3 to 30 days below the enterprise tier. For privacy-sensitive workloads that default is a real constraint.

EU residency is an enterprise deployment project

Portkey's managed service runs on its own cloud infrastructure, and no self-serve EU region is documented for it. Residency is addressed through enterprise hybrid or private-cloud deployment, where the gateway and data plane run in your environment. Where inference itself runs is a separate matter again, settled through your own contracts with each model provider.

The Opper Way

What Opper gives you that Portkey doesn't

A managed gateway where the models, the billing, and the privacy defaults come with the account.

Models come with the account, not with your key ring

Portkey is a control panel over provider accounts you already have. Opper is the account: 700+ models from Anthropic, OpenAI, Google, Mistral, and more, billed through one key with no markup on inference. There are no upstream credentials to wire, rotate, or reconcile, and bringing your own keys stays optional.

  • One key and one bill for every model
  • BYOK optional, never required
  • Automatic fallbacks across providers
See how pricing works
Portkey (bring your own keys)
  • · OpenAI account + key
  • · Anthropic account + key
  • · Google account + key
  • · Mistral account + key
  • · One bill per provider
  • · Rotate every key yourself
Opper (models included)
  • One API key
  • 700+ models ready to call
  • One pay-as-you-go bill
  • No markup on inference
  • BYOK optional
Models, billing, and keys in one account

Tracing when you opt in, metadata only by default

Opper's gateway meters cost, latency, and tokens per call and stores no prompt bodies by default. When you want full observability, the Control Plane adds span-level tracing, evaluations, and audit logs as an explicit opt-in, so recording is a decision you make rather than the resting state of the platform.

  • No prompts stored by default
  • Span-level tracing on opt-in
  • Searchable, exportable audit logs
See observability features
Portkey log, default
  • prompt_body: stored
  • response_body: stored
  • metadata: stored
  • retention: 3-30 days
Opper log, default
  • prompt_body: not stored
  • response_body: not stored
  • cost + latency: metered
  • tracing: opt-in
Recording is a choice, not the resting state

Self-serve EU residency

Run the same Anthropic, OpenAI, Google, and Mistral models on EU infrastructure, AWS Bedrock Frankfurt, Azure EU, or Berget AI, from your first call. No hybrid data plane to stand up, no private cloud to operate, and no enterprise contract before the first request.

  • EU regions for major model families
  • GDPR-aligned subprocessor chain
  • One-line region pinning
Explore the LLM Gateway
Same model, EU-hosted by default
US route
claude-sonnet-4.5
Anthropic US
EU route
claude-sonnet-4.5
AWS Bedrock Frankfurt
GDPR-compliant by default
Data residency contracted with EU subprocessors

Same OpenAI SDK, change one line

Drop-in compatible with the OpenAI client, the same integration path Portkey uses. Point base_url at Opper, use any of 700+ models from any provider, and keep your existing code, prompts, and tooling. No new SDK to learn.

  • 700+ models
  • Zero rewrites
  • Automatic fallbacks
from openai import OpenAI

client = OpenAI(
  base_url="https://api.opper.ai/v3/compat",
  api_key=OPPER_API_KEY,
)

response = client.chat.completions.create(
  model="openai/gpt-4o-mini",
  messages=[...]
)

# Same SDK, 700+ models, automatic fallbacks

An agent platform, not a security checkpoint

Prisma AIRS positions the gateway as a centralized control point to monitor and govern enterprise AI traffic. Opper points the other way: the Agent SDK, Agent CLI, and AI Wallet turn the same unified API into a platform for building, shipping, and monetizing production agents.

  • Agent SDK in Python and TypeScript
  • Agent CLI for Claude Code, OpenCode, Codex
  • User-funded inference via AI Wallet
Just a router
Proxy → upstream
Opper
Agent SDK

Build headless agents in Python or TypeScript

Agent CLI

Launch Claude Code, Codex, OpenCode on any model

AI Wallet

User-funded inference, no billing UI to build

Control Plane

Observe, route, steer, guard, comply

One platform, one API key

Model catalog & pricing

Access the latest models from leading providers with unified, transparent pricing per million tokens.

All prices are per 1M tokens • EU and US regions available • Prices subject to change, see docs for latest

ProviderModelRegionInput (1M tokens)Output (1M tokens)
Loading models...

Custom Models & BYOK

Bring your own API keys or add custom model deployments using the Opper CLI or API.

opper models create example/my-gpt5 azure/gpt5-production YOUR_API_KEY

Looking for a specific model? View complete model list →

FAQ

Opper vs Portkey FAQ

What happened to Portkey?

+
Palo Alto Networks announced its intent to acquire Portkey on April 30, 2026 and completed the acquisition on May 29, 2026. Portkey's technology now forms the foundation of the Prisma AIRS AI Gateway, which became generally available on July 16, 2026. The standalone portkey.ai product and pricing page remain live, and what the long-term roadmap looks like for the standalone offering is not stated.

What is the difference between Opper and Portkey (Prisma AIRS AI Gateway)?

+
Both are AI gateways with OpenAI-compatible APIs, tracing, and guardrails. Portkey is a bring-your-own-keys control panel that records requests and responses as its core product, now aimed at enterprise security buyers as part of Prisma AIRS. Opper is a managed gateway where models, billing, and EU data residency come with the account, 700+ models through one key, with no prompts stored by default.

Is Opper a drop-in replacement for Portkey?

+
For the gateway path, yes. Both expose OpenAI-compatible endpoints, so you swap base_url and API key on the OpenAI SDK and keep your prompts, tool calls, and streaming code. The practical difference is setup: on Opper there are no provider keys to wire up because models are billed through your Opper account, and bringing your own keys stays optional.

Does Opper store my prompts the way Portkey records logs?

+
No. Portkey's observability records requests and responses by default, with log retention windows of 3 to 30 days below the enterprise tier. Opper stores metadata only by default, per-call cost, latency, and token counts, and full tracing is an explicit opt-in through the Control Plane.

Does Opper offer EU data residency without an enterprise contract?

+
Yes, EU residency is self-serve from your first call: run the same Anthropic, OpenAI, Google, and Mistral models on AWS Bedrock Frankfurt, Azure EU, or Berget AI with no plan upgrade. Portkey's public documentation describes no self-serve EU region for its managed service, residency is addressed through enterprise hybrid or private-cloud deployment. And since Portkey routes inference through your own provider keys, where the models themselves run is a separate question you settle with each provider.

When is Portkey or Prisma AIRS the better choice?

+
Portkey is the honest pick when you need self-hosted or in-VPC deployment beyond what Opper's enterprise plans cover, since its open-source gateway core and hybrid data plane give you more deployment control, when you want a full prompt-management CMS with templates, versioning, and a playground, or when your organization is consolidating on Palo Alto Networks and wants the gateway inside the same security platform. It also states compliance with SOC 2, ISO 27001, GDPR, and HIPAA, and Opper does not currently publish SOC 2 or ISO 27001 certifications. Its observability is among the deepest in the category if recording requests and responses fits your data-handling rules.
Run Claude Code, Codex, OpenCode on any model

Agent CLI

Run Claude Code, Codex, OpenCode on any model

The same OpenAI-compatible gateway also powers the Opper Agent CLI: install once, log in once, and route Claude Code or any other coding agent through 700+ models on a single key.

The Portkey alternative with models included

Get an API key and ship in minutes, with no provider keys to wire, no prompts stored by default, and EU residency from the first call.

Get startedView Documentation