Fast connection
Set the UVOX endpoint and gateway key.
Connect Claude Code through UVOX, keep your Anthropic account under your control and select Auto, Maximum Savings or Zero-Loss Cache from the dashboard.
Combined modes are beta and can change the provider-bound context. Zero-Loss Cache preserves complete request content.
ANTHROPIC_BASE_URL=https://api.uvox.tech ANTHROPIC_AUTH_TOKEN=YOUR_UVOX_KEY ANTHROPIC_API_KEY= claude
UVOX supports demanding AI development use cases with simple deployment and measurable economics.
Set the UVOX endpoint and gateway key.
Continue using Claude Code and your preferred development environment.
Monitor context avoided, cache-write input, cache-read input, provider cost, route and fallback status.
Exact-copy, structured-output, high-risk and tool-history requests automatically use complete-context fallback when configured.
Create a UVOX key, connect your Anthropic provider key securely and launch Claude Code with the UVOX endpoint.
set "ANTHROPIC_BASE_URL=https://api.uvox.tech" set "ANTHROPIC_AUTH_TOKEN=YOUR_UVOX_KEY" set "ANTHROPIC_API_KEY=" claude
Test your own workload before moving important production traffic. Results vary and are not guaranteed.
X-UVOX-Mode. Combined beta falls back to the complete original request when its safety gate does not pass.Auto and Maximum Savings coordinate context optimisation with explicit Anthropic prompt caching. Zero-Loss Cache keeps complete request content.
Uses the combined route only when the request passes configured safety and confidence checks.
Uses the tenant-configured beta confidence and reduction limits while retaining protected content and fallback.
Preserves every request message and applies supported Anthropic cache-control markers.
UVOX publishes integration instructions, security practices and measured results—not the internal optimisation logic.
Use the free monthly allowance and evaluate your own development workload.
Create free accountUse your own Anthropic and UVOX keys to compare direct cold/warm requests with UVOX cold/warm requests. Verify prompt-content hashes, cache-write/read tokens, provider/model identity and fresh response IDs.