AI agent CLI for retail operations. Manage inventory, orders, customers, and analytics from your terminal — or let Claude, Gemini, or any MCP client do it for you.
Each tool is loaded dynamically from your database. Ask Claude to do any of this in natural language.
Sales analytics with flexible date ranges
Adjust quantities, set stock, transfer between locations
Find, create, update customer records
Find orders, get details, manage purchase orders
Find, create, update products and pricing
Create, approve, receive, cancel POs
Move inventory between locations
Send emails, manage inbox
Low stock and pending order alerts
Generate COAs and documents
Find and list suppliers
View audit logs for any operation
Install the package, set your credentials, add it to Claude.
1# Install2npm install -g @whale-code/cli3# Add to Claude Code (~/.claude/settings.json)5{6"mcpServers": {7"whale": {8"command": "whale-code",9"env": {10"SUPABASE_URL": "https://your-project.supabase.co",11"SUPABASE_SERVICE_ROLE_KEY": "your-key",12"DEFAULT_STORE_ID": "your-store-uuid"13}14}15}16}
1# Or for Claude Desktop2{3"mcpServers": {4"whale": {5"command": "npx",6"args": ["@whale-code/cli"],7"env": {8"SUPABASE_URL": "https://your-project.supabase.co",9"SUPABASE_SERVICE_ROLE_KEY": "your-key",10"DEFAULT_STORE_ID": "your-store-uuid"11}12}13}14}
Its own AI agent. Run whale and you get an interactive terminal with the full agent loop — reads/writes files, runs commands, manages inventory, processes orders. Same workflow as Claude Code, but purpose-built for retail ops.
The same tools are exposed as an MCP server. Add it to Claude Code, Claude Desktop, Gemini, Cursor — any client that speaks MCP gets full access to your inventory, orders, customers, and analytics.
Tool definitions live in ai_tool_registry. Add a row, the MCP server picks it up on next refresh. No rebuild.
whale code has its own agent loop, but the MCP server works with any client. Same tools, same backend, different frontends.
The CLI runs on your machine. File access, command execution, code editing — all local. Server tools proxy through edge functions to your database.
Every action — CLI or MCP — is logged with W3C trace context. See exactly what happened, when, and how much it cost.
W3C trace_id + span_id on every call. Correlate across CLI sessions, MCP calls, and edge functions.
Input/output tokens and dollar cost logged per turn. Track spend across conversations and agents.
Every conversation gets a UUID. Every turn is numbered. Full replay from audit_logs.
Telemetry never blocks the agent. Async inserts to audit_logs — zero impact on response time.
We build the same things we use — and we'll build them for you.
Next.js web apps, storefronts, admin dashboards, landing pages. Deployed and maintained.
Custom AI agents for your workflow. MCP servers that connect Claude, Gemini, or any client to your backend.
Database design, edge functions, auth, CI/CD. We set it up and keep it running.
Telemetry pipelines, audit logs, cost tracking, alerting. Know what your AI is doing and what it costs.