Start here

What is Panicly?

A spend firewall and control layer for AI provider traffic
View as Markdown

Panicly sits between your application and upstream AI providers. It authenticates project traffic, evaluates policy, records the decision, and forwards only approved requests.

One-line model

Panicly is a gateway and control layer that enforces policy before provider spend happens.

Control before cost

Requests pass through authentication, quota, and policy checks before reaching providers such as OpenAI, Anthropic, OpenRouter, Google, or Vercel.

Operator evidence

The dashboard centers request decisions, Workspace Audit, usage, credits, and forensics.

Emergency shutdown

Kill Switch gives operators a direct hold control for protected project traffic.

What Panicly is not

Do not position Panicly as a model marketplace

The product notes distinguish Panicly from OpenRouter: Panicly is about protection, limits, policy, evidence, and spend control. OpenRouter is about model discovery and routing breadth.

Panicly can forward to provider-compatible endpoints, but its primary value is not “more models.” Its value is making model traffic controllable.

Core product story

Panicly request flow from client app to gateway, controls, provider, and evidence ledger
Panicly evaluates the expensive path before the provider sees it.

The public story should stay grounded in five claims:

  • Requests are authenticated with a Panicly project key.
  • Rules and plan state are checked before provider forwarding.
  • Only approved traffic reaches the upstream provider.
  • Decision metadata is logged for later inspection.
  • Provider credentials are stored as encrypted project secrets.

Where to go next

Send your first request

Create a project, connect a provider, generate a Panicly key, and route a first call.

Understand the gateway

See the full request lifecycle from key authentication through policy and logging.