What we are building.
Our public backlog — what's live, what's in progress, and what's coming next. Updated continuously as we ship.
One base URL, one auth token. Drop-in compatible with the OpenAI request shape — no SDK changes required.
Per-request model scoring across cost, latency, quality, and capability signals to always pick the best fit.
When any upstream provider returns a 5xx or timeout, JTRouter retries on the next-best model with no client error.
Pay your balance in USDC, SOL, ETH, or JTVO. Invoices settle on-chain within seconds.
Claude, GPT, Gemini, DeepSeek, Grok, Mistral, and more — all accessible through a single endpoint.
Full per-request logs with latency, cost, model used, fallback path, and token breakdown.
Set per-member or per-project spend caps. Alerts fire at 80% and hard-stop at 100% — no surprise overages.
Push request events, quota alerts, and settlement confirmations to your own endpoint in real time.
Richer breakdown charts: cost by model, by team member, by prompt template, and by time-of-day heatmap.
Set a monthly budget and let JTRouter auto-route across DeepSeek, Claude, and GPT to stay under it without sacrificing quality.
First-party typed SDKs with built-in retries, streaming helpers, and model selection hints.
Run the JTRouter gateway inside your own VPC. Same routing logic, zero data leaves your network.