Getting started

Neonjelly is ecommerce intelligence for agents. An agent calls tools over HTTP and gets store, product, app, mover, change-feed, and Signal data back as JSON. Catalog feeds track product / traffic / revenue moves. Signals (paid only) watch a competitor for ads, units, and inventory.

What exists today:

• Start: /start — one click mints a trial, runs a live catalog hit, opens Cursor
• MCP: https://mcp.neonjelly.io/mcp (needs a key or connect URL)
• Connect URL: https://mcp.neonjelly.io/c/<njc_…>/mcp (auth in the path)
• 60 tools — composed playbooks plus catalog wrappers. List: /docs/tools
• Catalog REST: https://mcp.neonjelly.io/v1 — GET-only. Same nj_ / njc_ token. See /docs/api
• Playground: /playground — live curl + run
• Start an agent on whoami (quota) then list_playbooks (VC / competitor / dropship / outreach / merch)
• Trial: POST /v1/trial (one per device + AI client, 14 days, 100 calls/day, 20/min)
• Paid: Stripe Checkout on /pricing. Success page /dash shows the connect URL and (if new) the nj_ secret.

What does not exist:

• User accounts — the token is the identity. /dash is the post-checkout install screen, not a product console.
• Scrape-on-demand, probe, or delete watches from the catalog. create_signal is the one write: paid plans only, tight cap.

First calls

1. whoami — plan, used today, remaining, expiry (not metered).
2. list_playbooks — pick a persona.
3. resolve_store q=gruntstyle if you only have a brand. Do not invent a domain.
4. Then the composed tool for the job (diligence_pack, get_contacts, …), not five raw wrappers.

Missing catalog store → not_found. That is final. Neonjelly does not queue a scrape.