AgentKeyRoute Docs

Govern the AI websites your team already uses, and the API calls your apps make

AgentKeyRoute is an AI governance gateway. Your apps and agents call it instead of calling model vendors directly. AgentKeyRoute Guard — the browser extension — governs what people type into ChatGPT, Claude, Gemini, Copilot and Perplexity. Every request on either path is identified, checked for sensitive data, held to your policy and compliance frameworks, sent to an allowed model, and recorded in a tamper-evident audit log.

Your appsClaude Code, Cursor, chatbots, agents, your software
AgentKeyRoute gateway
  1. 1. Identify the app
  2. 2. Detect sensitive data
  3. 3. Apply protection
  4. 4. Choose a model
  5. 5. Check the answer
  6. 6. Record everything
ModelsOn-prem, your private cloud, or public APIs such as OpenAI, Anthropic and OpenRouter

Start here

What it solves

The situationWhat AgentKeyRoute does
An employee pastes a customer record into ChatGPT.com.AgentKeyRoute Guard sees the paste in the browser and blocks, redacts or logs it per your profile — or routes the whole session through your gateway on a company key.
A clinician pastes a patient note into a chatbot.Health data is detected and kept on your own models, or redacted before it reaches an outside provider.
Developers use Claude Code, Cursor or Codex on your codebase.Code may use approved providers, credentials are blocked, and every session is attributed to a person.
Auditors ask how you control AI.Export an evidence pack with framework scores, control checks and a verified audit chain.
Nobody knows which teams use which models, or what it costs.Every call is attributed to an app, team and person, with live cost and budgets.
A model provider has an outage.Requests fall back to the next allowed model automatically, within the same rules.
Someone tries to jailbreak your public assistant.Prompt-injection attempts are blocked and a safety model screens requests and answers.

Nothing changes in your apps

The gateway speaks the OpenAI and Anthropic APIs. To connect a tool you change two things: the address it calls and the key it uses. To govern a public AI website, you install AgentKeyRoute Guard and the admin picks the mode per site.

Address
OpenAI-compatible base URLhttps://gateway.agentkeyroute.com/v1
Anthropic base URLhttps://gateway.agentkeyroute.com
Consoleapp.agentkeyroute.com
On a dedicated or on-premises installation?
Use your own gateway address everywhere these docs show https://gateway.agentkeyroute.com. The console shows it under Connect an app → Developer reference.