Give this desk its own home.

No deployment connected yet.

Create a new Railway project from jaycarson373-tech/grok-desk. Attach a new volume at /data. Use one replica. Never reuse KATO’s backend, database or operator token.

Service → Variables

PORT=8080
DESK_DATA_DIR=/data
DESK_OPERATOR_TOKEN=<new private random token, at least 32 characters>

Generate a domain and set NEXT_PUBLIC_SITE_URL. Verify storage and authentication first. The signal engine and trading switches are listed in docs/RAILWAY.md. Real trades need a funded wallet and stay inside the frozen caps.

Model connections

The HEAD ledger report uses Grok through your own OpenRouter key: set both OPENROUTER_API_KEY and OPENROUTER_MODEL (for example x-ai/grok-4.3; any model with structured outputs works, such as openai/gpt-oss-120b for lower cost) on the server. Optional OPENROUTER_FALLBACK_MODELS lists up to three backup models. Requests ask OpenRouter for zero-data-retention providers only. It drafts a summary only, with no tools, search or execution, and has not been live-tested until a key is configured. Without OpenRouter, the inherited optional OpenAI helper is the fallback. The model choice only changes the wording of the HEAD summary; the baseline signal engine is rules-only. JUDGE remains Jev. Neither model is active without its own credentials, and Jev also needs complete sourced evidence. Broader Grok research is still planned.

Token launch, creator fee collection and buyback/burn are not implemented. See docs/RAILWAY.md in the repository for deployment notes.