Common questions about jusCode.

Short, specific answers. Each answer is also formatted as schema.org FAQPage JSON-LD so AI answer engines and Google can cite them directly.

• What is jusCode?

  jusCode is an OpenAI-compatible inference gateway at https://api.juscode.co that routes each request to the cheapest capable model for the task. It's designed for coding agents (Claude Code, OpenCode, Cursor, Aider, Cline, Continue, Goose, and any custom OpenAI-compatible harness) and reduces typical inference costs 50–80% with no workflow change.

• How is jusCode different from OpenRouter?

  OpenRouter is an aggregator: you pick the model per call, they route to the provider. jusCode is a smart router: we pick the model per call, optimized for coding-task quality and cost. OpenRouter is best when you already know which model you want. jusCode is best when you'd rather have the system decide.

• Does jusCode work with Claude Code?

  Yes. Set ANTHROPIC_BASE_URL=https://api.juscode.co and ANTHROPIC_AUTH_TOKEN=jcg_... in your shell. Claude Code's OpenAI-compatible mode routes every call through jusCode without other workflow changes. Full guide at https://juscode.co/docs/claude-code/.

• Does jusCode work with Cursor?

  Yes. In Cursor Settings → Models, set the Custom OpenAI Base URL to https://api.juscode.co/v1 and the API key to your jcg_ token. Add jusCode-auto to Custom Models. Cursor calls route through jusCode for 60–80% savings. Detailed guide at https://juscode.co/blog/cursor-too-expensive-options/.

• Is jusCode OpenAI-compatible?

  Yes. Both the Chat Completions API (/v1/chat/completions) and the newer Responses API (/v1/responses) are supported with the standard OpenAI request shape. Streaming, tool calls, vision inputs, JSON mode, and parallel tool calls all work. Any client built for OpenAI works by changing base_url and api_key only.

• How much does jusCode cost?

  Pay-as-you-go credits, $5/$10/$25/$50 packs via Stripe. No subscription required for solo accounts. Teams with more than one member need a $1/seat/month subscription (the first user is always free). Per-token rates depend on the underlying model selected per call. Typical blended cost for coding workloads is $0.20–$1.00 per million tokens. Credits never expire.

• Is there a free tier?

  Every new signup gets $0.05 in starter credit: enough to test the full integration end to end and run a few dozen real coding-agent tasks before topping up. There is no time-based free trial; credits don't expire.

• Can I use jusCode in production?

  Yes. The gateway runs on Cloudflare Workers with global edge deployment. Per-key and per-tenant rate limits, per-user soft caps, anomaly alerts, and margin-drift monitoring are built in. We do not yet offer a formal SLA or enterprise tier. If you need either, email hello@juscode.co.

• What models does jusCode use?

  We route across a curated pool of frontier and open-weights models from Anthropic, OpenAI, Together, Fireworks, and Cloudflare Workers AI. The specific model per call is selected based on task type, context length, and cost. Model identity isn't surfaced. That opacity is part of how the routing delivers its cost. Your Usage tab shows per-call cost so the bill stays transparent.

• Do you store my prompts or generated code?

  We store token counts, model selection, and timestamps for billing and anomaly detection. We do not store prompt content, completion content, or generated code beyond a short transient window required to forward the request to the upstream provider. Upstream providers (Anthropic, OpenAI, etc.) have their own retention policies; see their docs.

• Where is jusCode hosted?

  Cloudflare Workers: global edge deployment with regional routing based on caller location. Data (auth tokens, usage records, pre-orders) lives in Cloudflare D1 in a single primary region. The site (juscode.co) is on Firebase Hosting; the API (api.juscode.co) is on Cloudflare Workers.

• How do I migrate from OpenAI / Anthropic / OpenRouter direct?

  Change two values in your client config: base_url to https://api.juscode.co/v1 and api_key to your jcg_ token (mint at https://juscode.co/developer). No other code changes needed. Existing model-name overrides are honored. Pass nousresearch/hermes-4-405b or anthropic/claude-sonnet-4.5 and we route accordingly.

• What if a model is temporarily unavailable?

  jusCode retries transparently across providers. If Anthropic Sonnet 4.5 is rate-limited, we fall back to an equivalent-tier model (GPT-5 or a top-tier open-weights option) automatically. Your client sees a successful response, not a 5xx error. Failure is only surfaced when all candidates fail.

• Can I see which underlying model was used for a call?

  No. Model identity stays ours. It's part of how the routing delivers its cost. The Usage tab in the dashboard (https://juscode.co/developer) shows per-call cost so you always know what you paid for, but not the routing decision behind it.

• Does jusCode work with Codex CLI?

  Not yet. Codex CLI integration is coming soon. Codex only speaks the OpenAI Responses API (`/v1/responses`), and our routing currently sends most traffic to Chat-Completions-only upstreams where Codex tool calls would silently drop. Rather than ship a half-working integration that quietly degrades on the cheap models, we're enhancing the gateway's Responses adapter to support full tool-call translation first. Background reading: https://juscode.co/blog/codex-cli-custom-provider/. Tell us when it's ready at https://juscode.co/preregister/?tier=codex.