blinkit-mcp
An API-only MCP server for Blinkit that enables automated shopping on the Indian quick-commerce platform, supporting login, search, cart, checkout, and payment via UPI.
README
blinkit-mcp
An API-only MCP server for Blinkit
(Indian quick-commerce). Every runtime call is a plain HTTPS request to blinkit.com — no browser
automation in the happy path. A browser is used only once, during research, to discover endpoints;
see RESEARCH.md.
How it works (the important bits)
- Cloudflare TLS fingerprinting blocks ordinary HTTP clients (curl, Node
fetch→ 403). We useimpitto impersonate Chrome's TLS handshake, so requests pass with no browser and no cookie/challenge. This is what makes API-only possible. - Bootstrap:
GET /v2/accounts/auth_key/with a fixedreq_keyconstant → deviceauth_key(no login). Sent on every call. - Login is headless OTP:
send_otp(phone)→verify_otp(phone, code)returns anaccess_tokenwe persist. No browser ever involved. - Server-driven UI responses are parsed defensively into clean product/order objects.
Setup
pnpm install
pnpm build
Add to your MCP client (e.g. Claude Code mcp config):
{
"mcpServers": {
"blinkit": { "command": "node", "args": ["/path/to/blinkit-mcp/dist/index.js"] }
}
}
State is stored in ~/.blinkit-mcp/:
session.json— device id, auth_key, access_token (secret), location, store. chmod 600.staples.json— your reorder catalog + scorer weights.cart.json— the current working cart.
Tools
Auth/login — blinkit_login_status, blinkit_send_otp, blinkit_verify_otp, blinkit_logout
Location — blinkit_set_location, blinkit_check_serviceability
Discovery — blinkit_search, blinkit_autosuggest, blinkit_pick_best, blinkit_recommendations, blinkit_home_feed
Cart — blinkit_add_to_cart, blinkit_remove_from_cart, blinkit_view_cart, blinkit_clear_cart
Reorder — blinkit_quick_reorder, blinkit_list_staples, blinkit_set_staple
Checkout/pay — blinkit_get_addresses, blinkit_checkout, blinkit_prepare_order, blinkit_pay_upi, blinkit_payment_status
Orders — blinkit_order_count, blinkit_order_history
Typical flow (fully headless; only PhonePe approval is manual)
blinkit_set_location { lat: 28.5653836, lon: 77.38265 } # once, persists
blinkit_send_otp { phone: "9XXXXXXXXX" }
blinkit_verify_otp { phone: "9XXXXXXXXX", code: "1234" } # stores access_token + phone
blinkit_pick_best { query: "milk", brands:["Amul"], attrs:["full cream"] } # → product
blinkit_get_addresses # → address_id
blinkit_checkout { items: [ <chosen> ], address_id } # → server cart_id (creates+binds+validates)
blinkit_pay_upi { cart_id, method:"collect", vpa:"name@ybl", wait:true }
# → pushes a UPI collect to PhonePe; streams status via notifications; you approve on your phone
Payment status & notifications
blinkit_pay_upiinitiates UPI payment (method:"qr"returns aupi://payintent link;method:"collect"+vpapushes a collect request to your UPI app). Withwait:trueit pollsverifyPaymentStatusuntil terminal and returnsfinal_status.blinkit_payment_statuschecks the last (or a given) in-flight payment on demand — returnspending/success/failed.- The server declares the MCP
loggingcapability and emitsnotifications/messageduring the wait ("waiting for approval", "✅ approved", "❌ failed") — surfaced live by Claude Code.
Payment is delegated to Zomato zpaykit; the human step is approving the UPI request in PhonePe. See
RESEARCH.md §8–9for the full captured flow.
Notes / limits
- Unofficial, reverse-engineered API — no stability guarantees; header/version constants may need bumping. Respect Blinkit's ToS and rate limits; intended for personal automation.
auth_key/req_keyconstants are overridable viaBLINKIT_REQ_KEYif Blinkit rotates them.- Not affiliated with or endorsed by Blinkit/Zomato. Use at your own risk.
License
MIT — see LICENSE.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。