Shadow AI

Your people already use AI. Govern it where they type.

Shadow AI is every ChatGPT, Claude, Gemini and Copilot tab that company data goes into without a policy in the way. OpenSmartRoute puts the policy on the device: a browser extension and a desktop app that check a prompt for personal data, secrets and price before it is sent, redact in place, and route the work that must stay in-house to your own models - with every decision on one audit trail.

Get a free API key Route this request in the playground

Free plan, no card. Open source under Apache-2.0; self-hosting is free forever.

From the benchmark

AI sites covered by the extension, plus every desktop app

ChatGPT, Claude, Gemini, Copilot and Perplexity through the browser extension; any application through the desktop app's command bar and writing assistant. Checks run on the device - the prompt is not sent to the platform to be analysed.

The whole benchmark, with the cases the router lost

The problem

Blocking the sites did not work. Nobody tells you which ones they use.

Finance pastes a forecast into a public chat tab, support pastes a customer's ticket, an engineer pastes a key. Each one is doing their job faster, and each one moves company data outside every control you have. A policy document and a blocked URL list change neither the behaviour nor the risk - they move it to a phone.

Personal data and secrets leave in prompts nobody reviews
Spend on personal AI subscriptions is invisible and unmanaged
When the auditor asks what went where, there is no record

How the router answers it

Three things it does for every request.

A check before the prompt leaves
The browser extension sits on ChatGPT, Claude, Gemini, Copilot and Perplexity; the desktop app works in any application. Each prompt is analysed on the device for personal data, secrets, injection and price - and redacted in place with one click - before it is sent anywhere.
The right destination, not a wall
Work that may use a public model goes there with the workspace's key and budget; work that carries personal data or must stay in region is routed to your own models or an internal agent. People keep a tool that works, and the data boundary holds.
One ledger for the organisation
Every routed request is metered and attributed to a member, a team and a project; the governance report lists the controls in force and the audit trail answers where a request went and which rule decided. Managed settings push the policy to every device.

Watch one decision

Illustrative: the lowest cost that still clears the quality the task needs. The live quote below is the real one.

A real request

The policy the extension and the desktop app enforce on every device, set once for the workspace.

The same policy applies in the browser, on the desktop and through the API, so the sanctioned path is also the easiest one.

Route it in the playground and read the trace

Exampleyaml

contains_pii: true            # treat every prompt as if it may carry personal data
data_boundary: private        # public models only when the prompt is clean
deny_targets: [consumer-chat] # never the unmanaged consumer endpoint
monthly_budget_usd: 400
log_payloads: redacted        # the ledger keeps the redacted prompt, not the original

Where this stands next to the alternatives, and how to go deeper.

Install

Install the extension from the Chrome Web Store or push it with a managed policy

Deploy the extension to the organisation

Or no install at all: the hosted API answers a plain curl with the decision, and the browser extension shows the price and the router's pick under the composer of the AI chat sites you already use.

Route your first request in under a minute.

OpenSmartRoute is open source and the free plan keeps the full trace of every decision. Create a workspace, mint a key and send the request above.

Create a free account Read the benchmark

Free plan, no card. Fifteen thousand decisions a month with the full trace.

Your people already use AI. Govern it where they type.

Blocking the sites did not work. Nobody tells you which ones they use.

contains_pii: true # treat every prompt as if it may carry personal data data_boundary: private # public models only when the prompt is clean deny_targets: [consumer-chat] # never the unmanaged consumer endpoint monthly_budget_usd: 400 log_payloads: redacted # the ledger keeps the redacted prompt, not the original

Your people already use AI. Govern it where they type.

Blocking the sites did not work. Nobody tells you which ones they use.

Three things it does for every request.

A check before the prompt leaves

The right destination, not a wall

One ledger for the organisation

The policy the extension and the desktop app enforce on every device, set once for the workspace.

Where this stands next to the alternatives, and how to go deeper.

Route your first request in under a minute.

Your people already use AI. Govern it where they type.

Blocking the sites did not work. Nobody tells you which ones they use.

Three things it does for every request.

A check before the prompt leaves

The right destination, not a wall

One ledger for the organisation

The policy the extension and the desktop app enforce on every device, set once for the workspace.

Where this stands next to the alternatives, and how to go deeper.

Route your first request in under a minute.