# TrustedRouter

> A privacy-first AI gateway that routes requests across 600+ models from 90+ providers through one OpenAI-compatible API, with an attested prompt path that never logs prompt or output content.

TrustedRouter is an AI routing gateway operated by Lore Hex Corp, a Delaware C Corporation founded by Joseph Perla. It provides developers and businesses with one OpenAI-compatible API for hundreds of models served across multiple providers, with routing controls for price, reliability, region, and privacy posture. The service runs through a hardware-attested gateway in GCP Confidential Space, and customers can cryptographically verify that the running code matches the published open-source repository.

## What It Is

TrustedRouter is a unified LLM routing API that sits between your application and model providers. It accepts standard OpenAI-shaped requests and routes them to the best available provider based on health, cost, privacy requirements, or geographic constraints. The core differentiator is a verifiable trust model: the gateway runs in a Trusted Execution Environment, publishes a live attestation record, and is designed so that even Lore Hex engineers cannot read prompt or output content. The homepage describes it as "an OpenRouter alternative for teams that need verifiable privacy, lower-cost model choices, provider fallback, and one API across hundreds of models."

## The Trust Architecture

The trust model is structural, not policy-based. Key design properties include:

- **Attested gateway**: The prompt path runs inside GCP Confidential Space. A hardware-signed measurement of the running binary can be compared to the public source commit and image digest.
- **No content logs**: Prompt and output content are intentionally excluded from the control plane and metadata systems. Usage logs contain metadata only.
- **Fail-closed behavior**: If attestation fails, or if a requested privacy floor (ZDR or confidential+E2EE) cannot be satisfied, the request fails rather than silently routing through an unverified fallback.
- **Separation of planes**: The production prompt path (`api.trustedrouter.com`) is architecturally separate from the dashboard and billing control plane (`trustedrouter.com`).

Customers can verify the trust evidence themselves by fetching a nonce-protected attestation token from the API and comparing the image digest to the published artifact on the trust page.

## Routing Model and Named Aliases

TrustedRouter exposes named routing aliases alongside direct model IDs:

- `trustedrouter/auto` — healthy provider rollover for general production traffic
- `trustedrouter/zdr` — hard zero-retention provider floor for legal and sensitive work
- `trustedrouter/e2e` — confidential compute + provider E2EE for the most sensitive workloads
- `trustedrouter/eu` — Europe-focused routing via the `api-europe-west4.quillrouter.com` endpoint
- `trustedrouter/synth` — multi-model synthesis presets (Iris, Prometheus, Zeus, and code-tuned variants)

Provider routing supports OpenRouter-compatible filters including `order`, `only`, `ignore`, `allow_fallbacks`, `min_privacy`, `data_collection`, `sort`, `max_price`, and `zdr`. Unknown options return a `400` error; recognized but unimplemented options return a stable `501`, never a silent no-op.

## Open-Source Deployment Model

The homepage states the product is "100% hosted open source." The control plane, website, gateway, infrastructure, and SDKs are all published in public repositories under the Lore-Hex GitHub organization. The primary repository (`quill-router`) is licensed under the Business Source License 1.1 — non-production use (security review, audit, local evaluation) is permitted without a commercial license, and each version converts to Apache License 2.0 four years after publication. SDKs are available for Python, TypeScript, Go, Rust, Swift, and Java, plus a Vercel AI SDK provider.

## Developer Setup Path

Migration requires changing one line: the `base_url` in any OpenAI-compatible SDK. The docs provide a 60-second quickstart, a migration guide from OpenRouter, agent setup instructions for Codex, Claude Code, Cursor, and Hermes, and an MCP server that lets agents inspect live models, pricing, provider posture, and credits. BYOK (bring your own provider key) is supported for teams with existing provider commitments or enterprise rate limits.

## Update: Active Development as of September 2026

The `quill-router` repository was last pushed on September 2, 2026, and the GitHub About page notes it was created in May 2026. The homepage references a feature in Axios dated August 25, 2026. The product is in active production with four live attested API regions across GCP, AWS, and Azure. The README documents a scale target of OpenRouter-class traffic (1 trillion tokens/day) and notes the current deployment does not yet meet that target — correctness, trust, billing, and SDK compatibility are described as in steady-state while capacity scales horizontally.

## Features
- OpenAI-compatible API (one base_url change to migrate)
- 600+ AI models from 90+ providers
- Attested gateway running in GCP Confidential Space
- Zero prompt or output content logging
- Fail-closed privacy filters (ZDR and confidential+E2EE)
- Named routing aliases (auto, zdr, e2e, eu, synth)
- Provider failover and health-based routing
- EU-focused regional routing
- Bring your own provider key (BYOK)
- Prepaid credits with auto-refill
- Stablecoin, card, ACH, and PayPal checkout
- MCP server for agent-readable catalog and docs
- Python, TypeScript, Go, Rust, Swift, and Java SDKs
- Vercel AI SDK provider
- Prompt caching support
- Batch API with encrypted result retention
- Video generation (Seedance, Veo, Sora, Runway, Kling, and more)
- Web search via Responses API with source citations
- Signed inference receipts for offline verification
- Request tagging for cost allocation
- Evals across providers through one API
- Live latency and availability benchmarks
- Public status page with per-provider health
- llms.txt and llms-full.txt for agent-readable docs
- Multi-model Synth presets (Iris, Prometheus, Zeus)
- User-provided model listing with 70% credit payout
- x402 stablecoin funding for agents
- Sign in with TrustedRouter for user-facing apps

## Integrations
OpenAI SDK, Anthropic SDK, Codex, Claude Code, Cursor, Hermes, Vercel AI SDK, Google Gemini, DeepSeek, Qwen, Mistral, Cerebras, Anthropic Claude, AWS Bedrock, Google Vertex AI, Azure OpenAI, Stripe, PayPal, PostHog, OTLP webhooks, GCP Confidential Space, Sentry, Google Cloud Spanner, Google Cloud Bigtable, MCP (Model Context Protocol)

## Platforms
WINDOWS, WEB, API, DEVELOPER_SDK, CLI

## Pricing
Freemium — Free tier available with paid upgrades

## Version
main (last pushed 2026-09-02)

## Links
- Website: https://trustedrouter.com
- Documentation: https://trustedrouter.com/docs
- Repository: https://github.com/Lore-Hex/quill-router
- EveryDev.ai: https://www.everydev.ai/tools/trustedrouter