TrustedRouter Provider
Adds TrustedRouter as an OpenAI-compatible chat and embedding provider. One key reaches OpenAI, Anthropic, Google, DeepSeek and others, with per-request routing and failover. Routing policies constrain which upstreams may serve a request - zdr for no-retention providers, e2e for confidential compute, eu for EU-hosted.
README
TrustedRouter Provider for Agent Zero
Adds TrustedRouter as a chat and embedding provider in Agent Zero. TrustedRouter is an OpenAI-compatible router: one API key reaches models from OpenAI, Anthropic, Google, DeepSeek, and others, with per-request routing and failover.
Install
Install from the Agent Zero plugin index, or manually:
mkdir -p usr/plugins/trustedrouter_provider
git clone https://github.com/Lore-Hex/agent-zero-trustedrouter.git usr/plugins/trustedrouter_provider
Then set your key — get one at trustedrouter.com/keys:
TRUSTEDROUTER_API_KEY=sk-tr-...
Provider ID trustedrouter maps to TRUSTEDROUTER_API_KEY automatically.
Use
Settings → Model Configuration → Chat Provider → TrustedRouter, then pick a model.
Model ids are namespaced — anthropic/claude-opus-4-7, openai/gpt-5.4-mini. A bare
gpt-5.4-mini is rejected.
Ids under trustedrouter/ are routing policies rather than single models. Each picks an
upstream per request and fails over if one is down:
| Model | Picks for |
|---|---|
trustedrouter/auto |
Capability, with failover |
trustedrouter/fast |
Latency |
trustedrouter/cheap |
Cost |
trustedrouter/zdr |
Providers that retain no data |
trustedrouter/e2e |
Confidential-compute providers |
trustedrouter/eu |
EU-hosted providers |
zdr, e2e, and eu are the useful ones for an agent that reads private files or
internal systems: they constrain which upstreams may serve the request at all, so the
constraint holds even as routing changes underneath.
Embeddings work the same way — select TrustedRouter as the embedding provider and
use an id such as openai/text-embedding-3-large or cohere/embed-v4.0.
The full catalog is at api.trustedrouter.com/v1/models.
License
MIT