🤖 SnipeWise for AI Agents

SnipeWise is an eBay auction-sniping service built for the agentic era. Give your AI assistant its own scoped API key and it can manage snipes, top up credit, subscribe, and leave reviews for you — money moves only with your explicit yes, never behind your back.

Connect your assistant in 60 seconds

In the SnipeWise dashboard, open API keys and create a key (give it only the scopes your assistant needs).
Copy the setup prompt below, paste it into your AI assistant (Claude, ChatGPT, Muse…), replacing skw_PASTE_YOUR_KEY_HERE with your key. The dashboard has a one-click Copy setup prompt button that fills it in for you.
That's it. Try: "Snipe eBay item 123456789 for me, max $50."

Setup prompt

What your agent can do

🎯 Manage snipesCreate, list, inspect, and cancel snipes. Set a max bid; SnipeWise fires it in eBay's final seconds.

🔍 Outcome checksOn-demand settle checks with plain-language result summaries (won / lost / pending / error).

💳 Billing, safelyTop up credit or subscribe via propose → confirm: the agent reads you a summary, you say yes, then it confirms.

🔔 Result webhooksHMAC-SHA256-signed push events (snipe.fired / settled / cancelled) so your agent always knows what happened.

⭐ Leave reviewsSubmit post-snipe reviews, including an AI Agent Friendly rating dimension.

🧾 Full audit trailEvery agent action is logged and reviewable — by you and by the agent itself.

Safety model least privilege

Scoped keys — snipes:read/write, billing:read/topup/subscribe, webhooks:write, reviews:write, account:read, keys:write, or *. Out-of-scope calls get 403 insufficient_scope naming the missing scope.
Propose → confirm for money — one-step charge endpoints reject API keys (403 use_agent_billing). The agent must read the plain-language summary to you verbatim and get an explicit yes before confirming with a single-use 15-minute token.
Daily money caps — each key is capped (default $100/24h, 0 = unlimited).
Idempotency — replay-safe money actions; double-submits never double-charge.
Machine-actionable errors — every agent error returns a stable code plus a next hint describing the corrective action.
Revoke anytime — one click in the dashboard kills a key instantly.

Key endpoints

Method & path	Does
`GET /api/snipes`	List your snipes
`POST /api/snipes`	Create a snipe {itemId, maxBid, …}
`GET /api/snipes/:id`	Snipe + plain-language `result` summary
`POST /api/snipes/:id/cancel`	Cancel a snipe
`POST /api/snipes/:id/recheck`	On-demand settle check vs eBay
`GET /api/item/lookup?itemId=…`	Look up an eBay item
`GET /api/billing/status`	Credit, subscription, `can_snipe`
`POST /api/agent/billing/<action>/propose`	Top-up / subscribe / trial / cancel → returns `summary` + `confirm_token`
`POST /api/agent/billing/<action>/confirm`	Execute after the user's explicit yes
`POST /api/agent/reviews/submit`	Submit a review {request_id\|snipe_id, rating, text, ratings?}
`GET /api/reviews/summary`	Public aggregate ratings no auth
`GET/POST /api/agent/webhooks`	Manage result webhooks
`GET /api/agent/actions`	Audit trail of this key's actions
`GET /api/agent/openapi.json`	Full machine-readable contract no auth

Rate limits

Scope	Limit
Auth endpoints (`/api/auth/*`)	10 requests / 15 min per IP
Agent key minting	30 / hour
Webhook registration	60 / hour
Admin auth	20 / 15 min per IP

Over-limit responses are 429 with code: "rate_limited", a Retry-After header (seconds), and X-RateLimit-Limit / X-RateLimit-Remaining headers.

Machine-readable resources

/api/agent/openapi.json — OpenAPI 3.1 contract (public)
/llms.txt — LLM-readable product summary
← Back to SnipeWise