Claude API Privacy — Through TrustedRouter
Call Anthropic Claude through a hardware-attested, open-source router. Anthropic's privacy posture plus a routing path you can verify.
Call Claude through a prompt path you can verify.
Anthropic's privacy posture is one of the best in the industry. But the path between your code and Anthropic still matters — and most routing layers are closed-source black boxes.
TrustedRouter routes to Claude (and 30+ other providers) through a hardware-attested, open-source gateway. Anthropic's privacy claims plus a verifiable router on top.
# Anthropic SDK
from anthropic import Anthropic
client = Anthropic(
base_url="https://api.trustedrouter.com",
api_key="sk-tr-v1-..."
)
# Or via OpenAI-compat SDK
from openai import OpenAI
client = OpenAI(
base_url="https://api.trustedrouter.com/v1",
api_key="sk-tr-v1-..."
)
resp = client.chat.completions.create(
model="anthropic/claude-sonnet-4.6",
messages=[{"role": "user", "content": "hi"}],
)TR doesn't store your prompts.
Attested gateway never writes request or response bodies. Verifiable in source. Anthropic's own privacy policy applies upstream.
Sonnet 4.6, Opus 4.7, Haiku 4.5.
Plus the older 3.x family, plus the full models from OpenAI, Google, Meta, Mistral, Z.AI, DeepSeek, Kimi, MiniMax. Same API, one key.
Hash-check the gateway image.
Fetch /attestation, grab the image_digest, compare to the published artifact. Now you know what code is between you and Anthropic.
You want Anthropic + something more.
Audit-driven routing. Your security team needs to read the router code. Anthropic's API is closed; TR's router is open.
Provider fallback. Anthropic has occasional API hiccups. TR's pool falls over to GPT, Gemini, or other Claude-class models on a hard failure.
Comparison playground. Try the same prompt across Claude, GPT, Gemini, Kimi side-by-side at trustedrouter.com/chat — no per-provider setup.
Unified billing. One TR contract, one invoice, one usage dashboard. No per-provider account sprawl.
Self-host. Run TR inside your own VPC for HIPAA, FedRAMP, or air-gapped environments. Anthropic-via-TR still works because the routing image is verifiable.
Current routes, prices, privacy, and measured performance.
Catalog facts come from the routes currently configured in TrustedRouter. Performance uses the same cached metadata snapshot as the public leaderboard. Prompts and outputs are not part of these measurements.
| Model | Providers | Context | Input | Output | Privacy | Measured route |
|---|---|---|---|---|---|---|
Claude Opus 5anthropic/claude-opus-5 |
|
1,000,000 | $5.25/1M | $26.25/1M | varies | 2092 ms TTFT anthropic · 76 tok/s · 100.00% available · n=8 |
Anthropic: Claude Opus 4.7anthropic/claude-opus-4.7 |
|
1,000,000 | $5.25/1M | $26.25/1M | varies | 2359 ms TTFT anthropic · 100.00% available · n=7 |
Anthropic: Claude Sonnet 4.5anthropic/claude-sonnet-4.5 |
|
1,000,000 | $3.15/1M | $15.75/1M | varies 2 cited scores | 2228 ms TTFT anthropic · 100.00% available · n=7 |
Anthropic: Claude Opus 4.8anthropic/claude-opus-4.8 |
|
1,000,000 | $5.25/1M | $26.25/1M | varies 5 cited scores | 1740 ms TTFT anthropic · 58 tok/s · 100.00% available · n=6 |
Anthropic: Claude Sonnet 5anthropic/claude-sonnet-5 |
|
1,000,000 | $2.1/1M | $10.5/1M to $15.75/1M | varies | 2871 ms TTFT anthropic · 100.00% available · n=6 |
Anthropic: Claude Opus 4.5anthropic/claude-opus-4.5 |
|
200,000 | $5.25/1M | $26.25/1M | varies | 2636 ms TTFT anthropic · 100.00% available · n=4 |
Browse every modelReview provider policiesOpen the full leaderboardSnapshot 2026-08-07T01:38:28Z