Taken from Portkey's own pricing page. Their plans are metered on recorded logs, not on tokens, and you bring your own provider API keys, so you pay OpenAI or Anthropic directly at their rates with no markup from Portkey.
| Plan | Price | Included | Retention |
|---|---|---|---|
| Developer | Free forever | 10k recorded logs/month. Portkey notes it is not suitable for production. | 3 days logs, 30 days metrics |
| Production | $49/month | 100k recorded logs/month, then $9 per additional 100k requests. Guardrails, RBAC, semantic caching. | 30 days logs, 90 days metrics |
| Enterprise | Custom | 10M+ logs/month, SSO, VPC and private cloud, SOC 2 Type 2, GDPR, HIPAA | Custom |
| Open source | Free, self-hosted | Universal API, routing, retries, fallbacks, guardrails, load balancing, basic dashboard | Your infrastructure |
Any comparison that opens by telling you to pay for something has already told you what it is. Before you buy a gateway, the honest list of things that cost nothing: Portkey's own Developer tier, Portkey's open source build running on your own infrastructure, LiteLLM, which is open source and self-hosted, and Cloudflare's AI Gateway, which has a free tier. Several of these do the core job. If routing, retries and fallbacks are all you need, start free and only pay when you hit a wall you can name.
✓ You are running production traffic and need more than 3 days of log retention
✓ You want guardrails, role-based access control and service account keys without building them
✓ You want prompt templates and versioning managed outside your codebase
✓ You do not want to run and patch a gateway yourself, and $49 a month is cheaper than the hours that takes
✓ You are pre-production or experimenting. The free tier covers you
✓ You already run your own infrastructure and would rather self-host. LiteLLM or Portkey's open source build
✓ Your real problem is a bill you cannot explain, not the routing layer. See below
Gateways are built for engineers. They answer which model handled a request, how long it took, and what it cost in aggregate. That is the right shape for a product team.
It is the wrong shape for an agency. If you are billing eleven clients and running AI work inside deliverables for all of them, the question you have to answer at the end of the month is not what your OpenAI bill was. It is which client is eating your margin. That requires tying every call to a client, a deliverable and a rate card, and then putting it somewhere your finance process can read. A gateway gives you the raw usage. It does not give you the attribution, and it does not put it on an invoice.
That is the part I build. Not a gateway, and not a product with a monthly fee. A cost attribution and reporting layer that sits on top of whichever gateway you already chose, wired into how you actually bill. If you keep Portkey, good. This works alongside it.
Not two gateways at once, no. Both sit between your application and the providers, so pick one. You can reasonably run the free tier in development and a paid plan in production, and you can put reporting on top of either without changing the gateway.
If your AI bill is fine but your client margins are not, that is a different problem. Worth 30 minutes.
Talk through your AI costs