Privacy & Data Architecture

How ToolRoute handles your data — short version: we don't touch it.

Recommendation only — never a proxy

ToolRoute tells your agent which model or tool to use. Your agent then calls the model directly with its own API keys. We never see your prompts, outputs, API keys, or tokens. We never proxy, intercept, or store any data flowing between your agent and the model provider.

What we collect

✓

Routing decisions

Which tier and model was recommended for a task category (not the task content itself).

✓

Voluntary telemetry

Agents can optionally report outcomes (success/failure, latency, cost) to earn routing credits and improve recommendations for everyone.

✓

Agent identity (optional)

Agents can register with a name to track credits and reputation. No email, password, or personal info required.

What we never collect

✗

Prompts, completions, or any LLM input/output

✗

API keys, tokens, or credentials of any kind

✗

Personal information — no emails, names, or accounts

✗

IP addresses or browser fingerprints

Data flow

1. Route

Agent asks "which model?" — we return a recommendation

2. Execute

Agent calls the model directly — we're not in the loop

3. Report (optional)

Agent shares outcome — earns credits, improves routing

Open source

ToolRoute is open source on GitHub. You can inspect exactly what data is collected, how routing works, and how telemetry is processed. Every API endpoint, scoring formula, and database migration is public.

Questions about privacy or data handling? Open an issue on GitHub.