BlogFind the right AI gateway when fees, ownership, and control matter most.
Concentrate.ai
LLM Gateway
Change models, providers, add fallbacks, and more by running all AI usage through one platform.
Updated October 2026. Every fee below links to the vendor's own pricing page.
OpenRouter is the default way to try many models through one API. Fees are where teams start looking elsewhere. OpenRouter adds no markup on inference, but it charges 5.5% (minimum $0.80) on every card credit purchase, 5% on crypto, and 5% on BYOK usage above $25,000 a month. At $50,000 a month in token spend, the card fee alone comes to about $2,750 a month.
Ownership is the newer reason. Stripe agreed to acquire OpenRouter in August 2026. It says its brand and neutrality won't change, but teams sending sensitive traffic are re-checking who sits in their request path.
The third reason is control. Once AI runs in production, engineering needs failover, finance wants spend by team, and security asks what data went to which provider. Those are company problems, and a model marketplace isn't built to solve them.
Below are eight alternatives that are still being actively built, what each costs on top of tokens, and where each one falls short.
Three of the best-known names in this category changed hands this year. Check whether a tool is still being built before you migrate to it.
| When | What happened | What it means for you |
|---|---|---|
| March 2026 | Mintlify acquired Helicone. The product is in maintenance mode. | Security fixes and new models only, no new features. Plan a move off the proxy. |
| May 2026 | Palo Alto Networks acquired Portkey, now sold as Prisma AIRS AI Gateway. | Upgrades route through Palo Alto sales. The roadmap follows Prisma's enterprise buyers. |
| August 2026 | Stripe agreed to acquire OpenRouter. | OpenRouter says its brand, product and neutrality stay the same. Your gateway is now owned by a payments company. |
| Tool | Best for | Fee on top of tokens | Hosting |
|---|---|---|---|
| Concentrate AI | Companies running AI in production | None: no platform, card or BYOK fee | Managed (self-hosted option) |
| Vercel AI Gateway | Teams already on Vercel | No markup; card processing passed through; ZDR and allowlist add-ons | Managed |
| Cloudflare AI Gateway | Cloudflare Workers shops | 5% on Unified Billing credits |
Best for: companies that want OpenRouter's one-API convenience plus the controls finance and security will ask for.
Concentrate gives you one OpenAI-compatible API, plus an Anthropic-compatible /v1/messages endpoint, for 196 models across 20+ providers. Switching is a one-line change to your base URL: https://api.concentrate.ai/v1.
Limits: a smaller catalog than OpenRouter. No OpenTelemetry export yet (on the roadmap). Data residency is US-only today. No batch or flex API, so batch-heavy workloads may get better pricing elsewhere.
Pricing: Pay-as-you-go with $0 fees. Enterprise pricing is volume-based.
Best for: TypeScript and Next.js teams already paying for Vercel.
Vercel charges no markup and no platform fee on tokens, including BYOK. It has budgets at team, project, key and user level, plus OpenTelemetry trace drains.
Limits: card processing fees are yours unless you're on Enterprise invoicing. Team-wide ZDR and team-wide provider allowlisting each cost $0.10 per 1,000 requests and need a Pro or Enterprise plan. BYOK spend doesn't count toward any budget, and budgets are soft caps: the request that crosses the limit still completes.
Pricing: provider list price. Add-ons are metered separately.
Best for: teams already running on Cloudflare Workers.
Analytics, caching, rate limiting and basic DLP are free on every plan. It supports spend limits keyed on metadata, such as a team ID.
Limits: Unified Billing adds 5% to credit purchases. Gateways created after September 24, 2026 bill logs on Workers Logs pricing. You build team attribution yourself.
Best for: enterprises that want the gateway inside a Palo Alto Networks security stack.
Palo Alto Networks acquired Portkey in May 2026 and now sells it as Prisma AIRS AI Gateway. Its strengths are routing, guardrails, prompt management and detailed logs.
Limits: Production is $49/month for 100K logs, plus $9 per extra 100K. SSO, granular budgets, VPC hosting and BAAs are Enterprise-only. Expect the roadmap to keep shifting toward Palo Alto's enterprise buyers.
Best for: teams that want a marketplace like OpenRouter.
600+ models, routing policies, caching and fallbacks.
Limits: a 5% markup on pay-as-you-go. SSO, RBAC, audit logs and PII detection are Enterprise-only.
Best for: platform teams that want to run their own proxy.
LiteLLM is the most widely used open-source gateway. It's free to self-host, and Enterprise adds SSO, SCIM, audit logs and RBAC.
Limits: you run it, patch it and scale it. LiteLLM doesn't sell inference, so you still need provider accounts. Many teams pair it with a managed inference layer such as Concentrate.
Best for: teams that need pooled per-team budgets inside a broader AI platform.
Pro is $25 per user per month, with 20K requests per user included and $20 per extra 100K. It supports up to 10 teams and 10 budget rules.
Limits: per-seat pricing grows with headcount, and model costs are billed separately.
Best for: high-throughput teams that want a fast self-hosted gateway.
Open source under Apache 2.0, with failover, virtual keys and budgets.
Limits: self-hosted only. Clustering, guardrails and the MCP gateway are in the enterprise tier.
| Tool | Token markup | Card / credit fee | BYOK fee | Subscription | SSO |
|---|---|---|---|---|---|
| OpenRouter | 0% | 5.5% (min $0.80) | 5% above $25K/mo | None | Enterprise |
| Concentrate AI | 0% |
No. Its costs come from the 5.5% credit-purchase fee and the 5% BYOK fee above $25K a month.
For managed gateways, Concentrate and Vercel both charge no token markup and no BYOK fee. Concentrate also charges no card fee.
Yes, with Concentrate, Vercel, Cloudflare and OpenRouter. Check whether BYOK spend counts toward budgets. On Concentrate it does by default. On Vercel it doesn't.
Most OpenAI-compatible gateways only need a new base URL and API key. Model names may need a provider prefix.
Concentrate enforces ZDR per key or org. Vercel offers ZDR per request or team-wide, the team-wide version at $0.10 per 1,000 requests.
| Managed |
| Portkey (Prisma AIRS AI Gateway) | Enterprise AI security programs | $49/mo Production + log overages | Managed or open source |
| Requesty | OpenRouter-style marketplace | 5% markup on pay-as-you-go | Managed |
| LiteLLM | Teams that want to run their own proxy | Free OSS; Enterprise is custom | Self-hosted |
| TrueFoundry | Per-team budgets in a broader AI platform | $25/user/mo Pro + request overages | Managed or VPC |
| Bifrost | Low-overhead self-hosted gateway | Free (Apache 2.0); enterprise tier | Self-hosted |
| 0% |
| 0% |
| None |
| Included free |
| Vercel AI Gateway | 0% | Processing passed through | 0% | Vercel plan | Vercel plan |
| Cloudflare AI Gateway | 0% | 5% (Unified Billing) | 0% | Workers plan for logs | Cloudflare plan |
| Portkey / Prisma AIRS | 0% | None | None | $49/mo and up | Enterprise |
| Requesty | 5% | None | None | None | Enterprise |
| TrueFoundry | 0% | None | None | $25/user/mo | Pro |