Comparison

AI gateway vs AI control plane.

Vidai, LiteLLM Proxy, Portkey and Kong AI Gateway, side by side on governance, drop-in compatibility and performance.

1 · How the products compare

Cost control, policy and audit on every call.

Performance only matters if the boundary is doing the work your organisation actually needs: attributing every pound to a team in real time, stopping the runaway agent before the invoice, enforcing policy in the request path, and producing the audit trail your regulator asks for.

Where applicable: inside-the-perimeter, built-in and acts, not alerts answers are stronger

Vidai.ControlPlane
LiteLLM Proxy
Portkey
Kong AI Gateway
Category
Sovereign AI control plane
Self-hosted OSS gateway
SaaS gateway
API-gateway plugin
Deployment
Single binary, inside your VPC
Self-hosted (FastAPI / Docker / k8s)
SaaS, Portkey-hosted
Plugin on Kong gateway
Where requests travel
Inside your perimeter only
Inside your perimeter
Across Portkey's perimeter on every call
Inside (self-hosted); cloud (Konnect)
Audit-trail location
Your storage, your retention, in-path
Your storage
Portkey storage
Your storage
Cost attribution
Per team, key, app, model, in real time, on a dashboard
Per virtual key, with a DB
Per key, per workspace
Per consumer
Spend control (acts, not just alerts)
Hard budget breakers stop the next call when a cap is hit; auto-fallback to a cheaper model
Per-key budget limits with alerts
Spend limits with alerts
Rate limits (calls, not £/$)
Cost-based routing
Routes to the cheapest model that meets policy and quality
Configurable fallback chains
Provider failover, not cost-led
Plugin chain
Rate-card server (where prices live)
Ships with a public rate-card server by default; private rate-card server available for negotiated pricing
Pricing hard-coded in config files; updates require a rebuild
Vendor-managed catalogue
Per plugin
SCD2 rate history (priced correctly after a vendor price change)
Built-in; every call is priced against the rate active at that moment
Not provided
Not provided
Not provided
Policy enforcement (PII redact, region lock, model allow-list)
Built into the request path
Guardrails integrations
Add-on guardrails module
Plugin ecosystem
BYOM guardrails (plug in your own classifier / PII / jailbreak detector)
First-class: any model on the OpenAI-compatible interface plugs in and runs inline
Fixed integrations (Lakera, custom hooks)
Vendor's guardrail module
Per plugin
Regulatory mapping (EU AI Act, DORA, GDPR Art. 44, UK CTPR, FCA, ISO/IEC 42001)
Mapped explicitly. The audit trail IS the regulator's answer
Not provided
Not provided
Per plugin
Cross-provider
OpenAI, Anthropic, Vertex, Bedrock, Azure OpenAI, plus self-hosted Llama / Mistral
Same set plus the OpenAI-compatible long tail
Same set
Via plugins
Install footprint
Single 25 MB binary, multi-arch Docker image, no external dependencies
Python runtime + Postgres + Redis for full functionality (hundreds of MB once deps are resolved)
SaaS by default; self-hosted is a Node.js app with Postgres + Redis
Kong gateway + Nginx + Postgres + the AI plugin

Two beats decide this comparison. The cost beat: attribution in real time, hard budget breakers that act not alert, a rate-card server that survives a vendor price change. The sovereignty beat: where requests travel, where the audit trail lives, and whether regulatory mapping comes with the product. Vidai is the only entry where all of these answers stay inside your perimeter.

2 · Performance

Governed traffic at proxy-pace speed.

The benchmark below was run with the other gateways in minimal pass-through mode (no auth, no rate limits, no policy applied). Vidai was tested with production features active, with auth, API key validation, rate limiting and routing. Even doing the governance work, Vidai sustains higher throughput at lower latency.

Throughput, requests per second

Higher is better

Vidai (with governance)
LiteLLM (pass-through)
Portkey (pass-through)
vs LiteLLM
vs Portkey
Low load
1,713
131
495
13×
3.5×
Modest load
1,959
151
557
13×
3.5×
Sustained production load
2,526
152
656
17×
3.9×

p95 latency, milliseconds

Lower is better

Vidai (with governance)
LiteLLM (pass-through)
Portkey (pass-through)
LiteLLM × worse
Portkey × worse
Low load
8.5 ms
646 ms
92 ms
76×
11×
Modest load
7.8 ms
303 ms
68 ms
39×
Sustained production load
12.8 ms
5,009 ms
1,200 ms
391×
94×

Even doing more work, Vidai is roughly 15× the throughput of LiteLLM and 4× the throughput of Portkey at sustained production load. The full benchmark, the methodology and the source code are at /blog/rust-python-vidai.

3 · What changes in your application

Base URL change, not an SDK rewrite.

Most products in this category serve a "100+ providers" list via the OpenAI-compatible shape. Vidai supports the same long tail. The drop-in difference is the native SDK: Vidai is drop-in with the native Anthropic SDK, the native Google GenAI SDK, and the OpenAI SDK as-is.

"Yes" answers are stronger — fewer code changes

Vidai.ControlPlane
LiteLLM Proxy
Portkey
Kong
What changes in your application
Base URL + API key. Nothing else.
Base URL + key; non-OpenAI SDKs may need rewriting to OpenAI shape
Base URL + key; OpenAI shape primarily
Per-plugin config
OpenAI SDK works against any upstream
Yes
Yes
Yes
Via plugins
Anthropic SDK works against any upstream
Yes, including non-Anthropic upstreams (OpenAI, Vertex, Bedrock, Azure)
Limited
Limited
Limited
Google GenAI SDK works against any upstream
Yes, including non-Google upstreams
Limited
Limited
Limited
Self-hosted open-weight models (Llama, Mistral, your fine-tunes)
First-class, inside the VPC
Via OpenAI-compatible adapters
Routes to self-hosted endpoints; data still crosses Portkey
First-class

The full SDK × upstream matrix is in the documentation.

4 · Which one is right for you

Pick the boundary that matches your constraint.

Vidai. Your organisation faces data-residency or audit obligations, runs agent-pace traffic, and wants the same boundary doing cost, policy and audit on every call.
LiteLLM Proxy. Early stage, OSS-only, throughput is low, and the Python runtime ceiling isn't yet your binding constraint.
Portkey. No perimeter constraints, want the fastest SaaS time-to-first-request.
Kong AI Gateway. Existing Kong shop adding AI traffic to a gateway you already operate.

If your principal constraint is cost attribution and enforced spend at agent-pace, the dedicated read is /use-cases/control-ai-spend.

5 · Frequently asked

Common questions on this comparison.

What is the difference between an AI gateway and an AI control plane?

An AI gateway is an application-layer proxy: it routes a model call and emits a log line. An AI control plane is infrastructure: every request also passes through policy enforcement, real-time cost attribution and an audit trail recorded inside your own perimeter. The control plane is what regulated organisations reach for once gateway-shaped tools stop answering the audit, finance and sovereignty questions on their own.

Do I need an AI gateway or an AI control plane?

If a single team uses one provider, throughput is low, and nobody is asking the compliance question, an AI gateway is the right choice. You need a control plane when more than one team is using the same infrastructure with different budgets, a regulator or auditor is asking what the AI traffic did, or agent-pace traffic has broken the cost dashboard you set up for chat.

Are there self-hosted alternatives to LiteLLM Proxy?

Yes. LiteLLM Proxy itself is self-hosted (Python). Vidai is self-hosted (a single binary inside your VPC) and adds in-path policy enforcement, per-team cost attribution and an audit trail mapped to regulatory frameworks. Kong AI Gateway is self-hosted via Kong's plugin model. Portkey is SaaS by default; the self-hosted option still crosses Portkey's perimeter when telemetry is collected.

What is the best AI gateway for enterprise?

The honest answer is that the best AI gateway for enterprise is usually not a gateway; it is a control plane. Enterprises that start with a gateway typically rebuild it as something stricter once the audit, residency and multi-team cost questions arrive. The comparison on this page lays out the trade-offs.

Can I use Vidai with my existing Anthropic SDK or Google GenAI SDK?

Yes. Vidai is drop-in with the native Anthropic SDK, the native Google GenAI SDK and the OpenAI SDK as-is. Base URL and API key change, nothing else. The full SDK × upstream support matrix is in the documentation at docs.vidai.uk/server/client-integrations.

Run the boundary your auditor and CFO can actually use.

A 20-minute walkthrough on a real deployment. Cost, policy and audit, governed from inside your perimeter.