TrustedRouter keeps OpenRouter's one-endpoint access to 600+ models, then adds a gateway that attests to the exact build it runs and never logs prompts, an EU gateway on every account, and a 5.5% fee with no subscription.
TrustedRouter is a hosted, OpenAI-compatible AI gateway: one API key and endpoint for 600+ models from 100+ providers, with automatic failover.
It's the best OpenRouter alternative for teams that send sensitive data to AI models or answer to auditors. Its gateway attests to the exact build it runs and never logs prompts, its EU gateway is open to every account, and its 5.5% fee falls as low as 0.1% at enterprise scale. None of the other hosted gateways compared here publishes a hardware attestation.
Most tools sold as OpenRouter alternatives sit in front of provider accounts you already have. If you want what OpenRouter actually gives you, one wallet, one API key and 600+ models, TrustedRouter is the like-for-like replacement, with a gateway you can verify.
The rest each cover part of the job. Self-hosted proxies leave you to run, patch and secure the gateway. Observability tools record your traffic instead of keeping it private. Cloud platforms tie you to one provider's catalog and prices: the same open model can cost up to 75% less on TrustedRouter than on Amazon Bedrock.
“‘Trust us, we won’t peek’ was never an acceptable answer to give a client.”
Robot, Robot & Human, a legal AI platform and TrustedRouter customer. Read the case study
| Gateway | Type | Attested gateway | Platform fee |
|---|---|---|---|
| TrustedRouter | Hosted router; on-prem option | Yes | 5.5% max on text and embeddings; as low as 0.1% at enterprise scale |
| OpenRouter | Hosted router | No | 5.5% on credit purchases; 8% on Business |
| Requesty | Hosted router | No | 5% markup |
| Vercel AI Gateway | Hosted router | No | No markup; team-wide zero retention billed per request |
| Cloudflare AI Gateway | Edge gateway | No | 5% on Unified Billing credits |
| LiteLLM | Self-hosted proxy | No | None; you run it and pay each provider |
| Portkey (Palo Alto Networks) | Gateway for your own keys | No | Gateway plans plus your provider bills |
| Helicone (Mintlify) | Observability proxy | No | Observability plans plus your provider bills |
| Amazon Bedrock | Cloud platform | No | Cloud list price per model |
| Google Vertex AI | Cloud platform | No | Cloud list price; premium on open models |
| Direct provider APIs | One provider per account | Not applicable | Provider list price |
The gateway runs on confidential hardware that signs a statement of the exact build it's running. Every build is published and signed, and the running build never logs prompt or output content.
Your security team can verify the attestation at trust.trustedrouter.com without taking our word for it. SOC 2 Type II audit in progress.
Every route carries a privacy label: zero data retention, confidential zero retention, or no verified claim.
Set a privacy floor on a request and TrustedRouter refuses it rather than fall back to a weaker route. Metadata retention is published, and content-free diagnostics expire after 90 days.
Send EU traffic to TrustedRouter's EU gateway and authentication, policy checks, provider selection and streaming start in Europe.
Use trustedrouter/eu for an EU-focused provider set led by Mistral, or pin requests to providers you've approved: if none can serve a request, it fails instead of leaving your list. OpenRouter keeps in-region routing for its Business and Enterprise plans.
When providers compete for each request, the same open model costs less. On a workload of 100M input and 20M output tokens a month, gpt-oss-120b costs 75% less on TrustedRouter's cheapest route than on Amazon Bedrock, and DeepSeek V3.2 costs 64% less.
Enterprise rates bring the platform fee as low as 0.1%.
| Pay as you go | Enterprise | |
|---|---|---|
| Platform fee | 5.5% on top of provider prices, at most, on text and embedding models | Falls with volume, to as low as 0.1% |
| Subscription or seats | None | None |
| Getting started | 1 million free tokens, no card | Migration credits, by approval, for teams moving off OpenRouter |
| Your own provider keys | Supported | Supported |
OpenRouter charges 5.5% on credit purchases on its Standard plan and 8% on its Business plan, which is also the plan that unlocks EU in-region routing. On TrustedRouter, the EU gateway comes with the standard fee.
Amazon Bedrock sets one list price per model. On TrustedRouter, competing providers serve the same model, so you pay the going rate instead of one cloud's list price.
| Same model | Amazon Bedrock | Trusted | You save |
|---|---|---|---|
| gpt-oss-120b | $27.00 | $6.75 | 75% |
| DeepSeek V3.2 | $99.00 | $35.45 | 64% |
| Qwen3 235B (2507) | $39.60 | $21.10 | 47% |
100M input + 20M output tokens/month, priced Oct 19, 2026. Cheapest TrustedRouter routes, fees included; privacy and precision vary.
Moving requests that don't need a frontier model saves more. The same workload costs $800 on Claude Opus 5.5 and $103 on GLM 5.2 through a zero-retention route.
The model provider's terms say it won't store your prompts or outputs. It's a contractual promise. TrustedRouter labels which routes have one, and "provider": {"min_privacy": "zdr"} refuses any route that doesn't.
TrustedRouter's own gateway runs on confidential hardware that proves which published build is handling your request, and that build keeps no prompt or output logs. Attestation covers our hop, so the model provider's terms still apply.
The prompt stays encrypted until it reaches a provider's verified confidential-computing hardware, and the route must also be zero-retention. Select them with "min_privacy": "confidential" or the trustedrouter/e2e model. Available on open-weight models today.
A hosted router with a 5% markup and an EU endpoint in Frankfurt. Its privacy rests on policy; it publishes no gateway attestation that we could find.
TrustedRouter matches the hosted, EU-ready setup and adds a gateway you can verify, privacy floors that fail closed, and enterprise rates down to 0.1%.
A hosted gateway with no token markup, designed around Vercel's platform and AI SDK. Per-request zero retention is free; enforcing it across a whole team costs $0.10 per 1,000 requests.
TrustedRouter works the same from any stack, and its gateway is attested.
An analytics, caching and rate-limiting layer on Cloudflare's network. Unified Billing adds a 5% fee on credits, and full request payloads are logged by default.
TrustedRouter never logs prompt or output content, and attests to the build that handles them.
An MIT-licensed proxy you deploy, scale, patch and secure yourself, with a separate contract for every provider. In March 2026, two LiteLLM releases on PyPI were found to carry malware that stole cloud and CI credentials.
TrustedRouter runs the gateway for you, signs every build and attests the one running, and offers an Apache-2.0 on-prem gateway when you need one inside your network.
A gateway for your own provider keys, now part of Palo Alto Networks' Prisma AIRS platform after the acquisition closed in May 2026. You still sign up for and pay every provider separately.
TrustedRouter includes the catalog, billing and failover across 100+ providers, with privacy you can verify instead of configure.
An observability proxy built to record requests for tracing and cost analysis. Since Mintlify acquired it in March 2026, it has been in maintenance mode with no new features.
TrustedRouter is built for the opposite job: keeping prompts out of every log.
Model access inside one cloud account, at that cloud's prices. The same open model can cost up to four times as much on Bedrock as through TrustedRouter, and in-region Claude endpoints carry a 10% premium.
TrustedRouter routes across 100+ providers, Bedrock and Vertex included, from one key.
The fewest intermediaries and the most paperwork: one contract, key, invoice, rate limit and failover plan per provider.
TrustedRouter puts them behind one key and one bill, and fails over across providers automatically.
A policy says what a vendor intends. TrustedRouter publishes a hardware attestation of the exact build handling your requests, and that build doesn't log prompts.
A router that falls back silently can send data where you didn't intend. With a privacy floor set, TrustedRouter refuses the request instead.
On OpenRouter, in-region routing starts with the 8% Business plan. On TrustedRouter, the EU gateway is open to every account.
TrustedRouter starts at 5.5% with no subscription and falls as low as 0.1% at enterprise scale.
Stripe has agreed to buy OpenRouter, Palo Alto Networks owns Portkey, and Mintlify put Helicone into maintenance mode. TrustedRouter is independent.
| OpenRouter | Trusted | |
|---|---|---|
| API | OpenAI compatible | OpenAI compatible; OpenRouter routing fields accepted |
| Platform fee | 5.5% on credit purchases (Standard); 8% (Business) | 5.5% max on text and embeddings; as low as 0.1% at enterprise scale |
| EU routing | In-region routing on Business and Enterprise plans only | EU gateway on every account; EU-only providers by allowlist |
| Prompt and output logging | Off by default; opt in for a 1% discount | Never logged |
| Hardware attestation | No | Running build attested; every build signed |
| Gateway source | Closed | Source-available (BUSL-1.1); on-prem gateway Apache-2.0 |
| Ownership | Stripe agreed to acquire (announced Aug 19, 2026) | Independent |
Stripe agreed to acquire OpenRouter on August 19, 2026, in the largest acquisition in its history. OpenRouter said it would keep operating under the same name, product and roadmap. After the close, its pricing, data handling and terms will be up to Stripe.
If you'd rather not wait to find out, moving to TrustedRouter is a base URL and key change.
Keep the OpenAI SDK and swap the base URL and key. OpenRouter-style routing fields such as provider.order, provider.only and provider.ignore carry over, so most migrations are a config change.
No card required. Start with 1 million free tokens.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.trustedrouter.com/v1",
api_key=os.environ["TRUSTEDROUTER_API_KEY"],
)
response = client.chat.completions.create(
model="trustedrouter/auto",
messages=[{"role": "user", "content": "Hello"}],
)Facts about other companies come from independent reporting, security research, regulatory filings, price trackers and model-lab documentation. TrustedRouter facts link to our own documentation.
Catalog facts come from the routes currently configured in TrustedRouter.
| Model | Providers | Context | Input | Output | Privacy |
|---|---|---|---|---|---|
Anthropic: Claude Opus 4.8anthropic/claude-opus-4.8 |
|
1,000,000 | $5.275/1M | $26.375/1M | varies 5 cited scores |
OpenAI: GPT-5.5openai/gpt-5.5 |
|
1,050,000 | $5.275/1M | $31.65/1M | ZDR 3 cited scores |
Google: Gemini 3.5 Flashgoogle/gemini-3.5-flash |
|
1,048,576 | $1.5825/1M | $9.495/1M | ZDR |
MoonshotAI: Kimi K2.7 Codemoonshotai/kimi-k2.7-code |
+11
|
262,144 | $0.692291/1M to $1.31875/1M | $3.4815/1M to $4.82135/1M | ZDR 5 cited scores |
Z.ai: GLM 5.2z-ai/glm-5.2 |
+28
|
1,048,576 | $0.3587/1M to $2.361723/1M | $1.5825/1M to $6.819476/1M | Confidential + ZDR 4 cited scores |
MiniMax: MiniMax M3minimax/minimax-m3 |
+12
|
1,048,576 | $0.24265/1M to $0.633/1M | $1.0128/1M to $2.532/1M | ZDR 4 cited scores |
Browse every modelReview provider policiesSnapshot 2026-10-28T07:07:57.445Z
For most teams leaving OpenRouter, TrustedRouter. It keeps the same OpenAI-compatible, one-endpoint access to 600+ models from 100+ providers, and adds a hardware-attested gateway that never logs prompts, an EU gateway on every account, and a 5.5% fee that falls as low as 0.1% at enterprise scale.
TrustedRouter. Like OpenRouter, it's a hosted marketplace: one wallet of prepaid credits, one API key and one OpenAI-compatible endpoint for 600+ models from 100+ providers, with automatic failover and OpenRouter-style routing fields. Most other alternatives are gateways that sit in front of provider accounts you open and pay for separately.
TrustedRouter charges provider prices plus 5.5% on text and embedding models, with no subscription. OpenRouter charges 5.5% on credit purchases on its Standard plan and 8% on Business, the plan that includes EU in-region routing. TrustedRouter's fee falls with volume, to as low as 0.1% at enterprise scale.
TrustedRouter. Change the base URL to https://api.trustedrouter.com/v1 and swap the API key. The OpenAI SDK stays, and OpenRouter-style routing fields such as provider.order, provider.only and provider.ignore are accepted. Teams spending more than $100 a month can request migration credits, granted by approval.
TrustedRouter publishes a hardware attestation of the exact build handling your requests, which none of the other hosted gateways here offers, and that build never logs prompts or outputs. Each route is labeled for zero retention and confidential compute, and a privacy floor makes a request fail rather than fall back to a weaker route.
TrustedRouter runs an EU gateway that every account can use: authentication, policy checks, provider selection and streaming start in Europe. The trustedrouter/eu model routes to an EU-focused provider set led by Mistral, and you can restrict requests to providers you've approved so nothing routes outside your list. OpenRouter offers in-region routing only on its Business and Enterprise plans.
TrustedRouter gives new accounts 1 million free tokens with no card required, and there's no subscription after that: you pay provider prices plus a 5.5% fee on text and embedding models. Teams spending more than $100 a month on LLMs can request migration credits, granted by approval.
TrustedRouter. Its agent setup guide covers Codex, Claude Code, Cursor and eval runners, so your coding agent gets one key for 600+ models with automatic failover, including open models that cost far less than frontier list prices. Requests run through the same attested gateway, which never logs your code or prompts.
TrustedRouter offers an Apache-2.0 on-prem gateway that runs inside your network, and publishes the source of its hosted gateway and control plane under BUSL-1.1 for review. LiteLLM's open-source proxy is self-hosted, so you operate and secure it yourself.
Yes. Change the base URL and API key and your existing OpenAI SDK code runs through TrustedRouter. Test any advanced fields your app depends on, such as tool calls or streaming events.
Stripe agreed to acquire OpenRouter, announced August 19, 2026. OpenRouter said it would keep operating under the same name, product and roadmap.