Use TrustedRouter with Cline & Roo Code: Cheap, Private Models
Point Cline or Roo Code at TrustedRouter's OpenAI-compatible API: 220+ model routes, automatic fallback, and no prompt or output logs in an attested TEE.
Your coding agent streams your entire repo through its API provider, so pick one you can verify.
Cline and Roo Code both speak the OpenAI-compatible protocol, which makes TrustedRouter a three-field setup: base URL, API key, model id. That buys you 220+ model routes across 30+ providers, including Claude, GPT, Gemini, DeepSeek V4, Kimi K2.7, GLM-5.2, and Qwen 3.5, with automatic fallback to another provider when one goes down.
Privacy matters more for agents than for chat. Every agent turn ships file contents, diffs, and terminal output upstream. TrustedRouter never logs prompt or output content. Ordinary synchronous and streaming inference does not retain it. The opt-in Batch API temporarily retains enclave-encrypted artifacts for up to 30 days. The prompt path runs inside a hardware-attested confidential-compute gateway whose source is open and whose attestation you can check live.
# Choose "OpenAI Compatible" in Cline or Roo Code, then paste:
# Base URL: https://api.trustedrouter.com/v1
# API key: your TrustedRouter key
# Model ID: trustedrouter/cheap
# Sanity check from a terminal:
curl https://api.trustedrouter.com/v1/chat/completions \
-H "Authorization: Bearer $TRUSTEDROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"trustedrouter/cheap","messages":[{"role":"user","content":"ping"}]}'Agent loops burn tokens
An agent reads files, retries edits, and reruns tests, so its token volume dwarfs chat usage. Route to trustedrouter/cheap and it picks the cheapest capable route, inside a TEE. Pricing is a thin markup over provider list prices, listed per model at /pricing. The smart-cheap-fast chooser at /choose helps you commit to a fixed route instead.
Attestation beats a privacy policy
The prompt path runs through an open-source, hardware-attested gateway. The image digest is published, the attestation JWT is signed by the CPU vendor's root key, and it binds to your live TLS session, so it cannot be replayed. Routes carry privacy tiers: Open, Zero-Data-Retention, or full TEE, where even the provider cannot read the prompt. Details at /security.
Chinese open weights, without the vendor
DeepSeek, Kimi, Qwen, GLM, and MiniMax routes are served by non-Chinese hosting providers on attested infrastructure, with ZDR and TEE tiers available. Your codebase does not go to the model vendor. One finding from our own testing: the same GLM weights answered 60/60 sensitive questions through our enclave versus 30/60 via the vendor-hosted API. The filtering is host-applied.
Check the gateway yourself before your agent sends a single file.
Run curl against the attestation endpoint with a nonce and you get back a JWT signed by the CPU vendor's root key. The image digest matches the published open-source artifact, and the attestation binds to your live TLS session. Privacy you can check beats privacy you're promised. The full procedure is at /security.
The gateway handles agent-scale volume: a document-processing company scaled from zero to over a billion tokens per week through the attested gateway within three weeks of onboarding. Pay with prepaid credits, bring your own provider key, or use x402 pay-per-request with MetaMask sign-in.
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 |
|---|---|---|---|---|---|---|
Anthropic: Claude Opus 4.8anthropic/claude-opus-4.8 |
|
1,000,000 | $5.275/1M | $26.375/1M | varies 5 cited scores | 1538 ms TTFT anthropic · 91 tok/s · 100.00% available · n=3 |
OpenAI: GPT-5.5openai/gpt-5.5 |
|
1,050,000 | $5.275/1M | $31.65/1M | ZDR 3 cited scores | Warming up |
Google: Gemini 3.5 Flashgoogle/gemini-3.5-flash |
+1
|
1,048,576 | $1.5825/1M | $9.495/1M | ZDR | 1278 ms TTFT google-vertex · 107 tok/s · 100.00% available · n=7 |
MoonshotAI: Kimi K2.7 Codemoonshotai/kimi-k2.7-code |
+10
|
262,144 | $0.70685/1M to $1.13096/1M | $3.587/1M to $4.90575/1M | ZDR 5 cited scores | 2262 ms TTFT kimi · 68 tok/s · 100.00% available · n=26 |
Z.ai: GLM 5.2z-ai/glm-5.2 |
+24
|
1,048,576 | $0.7174/1M to $2.434307/1M | $1.5825/1M to $7.029062/1M | E2EE 4 cited scores | 1806 ms TTFT baseten · 128 tok/s · 100.00% available · n=145 |
MiniMax: MiniMax M3minimax/minimax-m3 |
+9
|
524,288 | $0.24265/1M to $0.633/1M | $1.0128/1M to $2.532/1M | ZDR 4 cited scores | 987 ms TTFT fireworks · 220 tok/s · 100.00% available · n=194 |
Browse every modelReview provider policiesOpen the full leaderboardSnapshot 2026-09-17T13:23:21.989Z
Questions
How do I set TrustedRouter as the API provider in Cline?
In Cline's settings, choose the OpenAI Compatible provider type, then paste three fields: the base URL https://api.trustedrouter.com/v1, your TrustedRouter API key, and a model id such as trustedrouter/cheap or any of the 220+ routes; the model chooser at trustedrouter.com/choose helps you pick one. No SDK or code changes are needed because the API is OpenAI-compatible. The same three fields work in Roo Code's OpenAI Compatible provider settings.
Which model routes are cheapest for coding agents?
trustedrouter/cheap automatically selects the cheapest capable route and runs it in a TEE, which suits high-volume agent loops. trustedrouter/fast optimizes for speed instead. Fixed routes for open models such as GLM glm-5.2-fast, DeepSeek V4 Flash, and Kimi K2.7 are listed with transparent per-model pricing, a thin markup over provider list prices, at trustedrouter.com/pricing.
Is my code private when an agent routes through TrustedRouter?
TrustedRouter never logs prompt or output content. Ordinary synchronous and streaming inference does not retain it. The opt-in Batch API temporarily retains enclave-encrypted artifacts for up to 30 days. The prompt path runs inside a hardware-attested confidential-compute gateway whose source is open source, and you can verify the attestation live against a published image digest. Each route also carries a privacy tier: Open, Zero-Data-Retention, or TEE, where even the serving provider cannot read your prompt.
Can I use DeepSeek, Kimi, or GLM in Cline without sending code to the model vendor?
Yes. DeepSeek, Kimi, Qwen, GLM, and MiniMax routes on TrustedRouter are served by non-Chinese hosting providers on attested infrastructure, with Zero-Data-Retention and TEE tiers available. Prompts do not go to the model vendor. You get the capability of the open weights while your codebase stays inside infrastructure you can verify.