apiguru-amazon-data
# Apiguru Agent Kit
Everything needed for an AI agent to discover, call and pay for the Apiguru
Amazon Data API — **with no account, no API key and no subscription**.
The existing backend is not modified by any of this. Not one line.
## Install in your agent
| Client | How |
|---|---|
| Claude Code | `claude mcp add --transport http apiguru https://mcp.apiguru.app/mcp` or `/plugin marketplace add apiguru-app/agent-kit` then `/plugin install apiguru@apiguru` |
| claude.ai, Claude Desktop, ChatGPT | add connector `https://mcp.apiguru.app/account` and sign in |
| Codex CLI | `codex mcp add apiguru --url https://mcp.apiguru.app/mcp` |
| Cursor | [Add to Cursor](cursor://anysphere.cursor-deeplink/mcp/install?name=apiguru&config=eyJjb21tYW5kIjogIm5weCIsICJhcmdzIjogWyIteSIsICJhcGlndXJ1LW1jcCJdfQ==) (runs `npx -y apiguru-mcp`) |
| VS Code | [Install in VS Code](vscode:mcp/install?%7B%22name%22%3A%20%22apiguru%22%2C%20%22command%22%3A%20%22npx%22%2C%20%22args%22%3A%20%5B%22-y%22%2C%20%22apiguru-mcp%22%5D%7D) |
| Gemini CLI | `gemini extensions install https://github.com/apiguru-app/agent-kit` |
| Hermes, OpenClaw, any stdio client | `npx -y apiguru-mcp` (npm, Node only) or `uvx apiguru-mcp` (PyPI) |
| Skill only (any agent) | `npx skills add apiguru-app/agent-kit` or ClawHub `apiguru-amazon-data` |
## Why this exists
The API is gated behind register → verify email → get key → top up. Every step
assumes a human. Agents cannot do any of it, so agent traffic bounces off the
front door.
And agents don't browse — they query indexes: the x402 Bazaar, the MCP
Registry, ClawHub. Apiguru was in none of them, and the repo had no OpenAPI
spec at all, so there was nothing machine-readable to publish.
## Telling us what is broken
Agents are the ones who find the defects in this API, so there are two ways
back to us and neither costs anything:
- **GitHub issues — preferred:**
<https://github.com/apiguru-app/agent-kit/issues>. A thread can hold a
reply, and a fix gets linked back to it.
- **The wall, for callers with no GitHub account:** one unauthenticated
`POST https://dash.apiguru.app/api/v1/feedback` with
`{"message": "...", "category": "bug|wish|praise|question|other",
"endpoint": "/search", "agent": "your-name/1.0", "contact": "optional"}`.
No key, no signup, never billed. Read it at
<https://dash.apiguru.app/feedback>, or `GET` the same URL for JSON.
- Over MCP: the free `send_feedback` tool. From the skill:
`python scripts/probe.py feedback --message "..." --category bug`.
This is not decoration. 1.1.3 exists because an agent reported that search
returned the brand as `product_title`, `null` for every
`product_num_ratings`, and delivery text with the words welded together.
## What is in this repository
| Path | What it is |
|---|---|
| `skill/apiguru-amazon-data/` | The agent skill: SKILL.md, references, and `scripts/probe.py` |
| `mcp/` | The MCP server published to PyPI as `apiguru-mcp` |
| `npm/` | The npm bridge, `apiguru-mcp` |
| `plugin/` + `.claude-plugin/` | The Claude Code plugin and its marketplace entry |
| `spec/` | `endpoints.json` and the generated `openapi.json` / `llms.txt` |
The hosted service that serves these endpoints is operated separately and is
not part of this repository.
## Licence
See `LICENSE`.
TDQS
Scored across 12 tools
Most tools target clearly distinct resources (products by keyword, single ASIN details, seller reviews, product reviews, offers, deals, best-sellers). The main overlaps are search vs seller_products (same filter surface, differing only in seller scope) and search's today_deals/deal_type filters vs the dedicated deals tool; descriptions flag these differences but an agent could still misselect.
Names are uniformly snake_case (product_details, seller_reviews, best_sellers, seller_profile_batch), which is predictable. Minor deviations: 'search' and 'deals' are bare verbs/nouns without a noun or verb complement, so the verb_noun pattern is not perfectly uniform.
12 tools is well within the ideal 3-15 band for an Amazon data API, and each earns its place: batch variants, the offers/reviews/seller splits, and the two free meta tools (list_capabilities, send_feedback) all serve distinct needs.
The surface covers product search, detailed records, reviews (product and seller), offers/stock, seller profiles and listings, best-sellers and deals, plus a capability-discovery tool. Gaps are minor: no explicit marketplace-list tool (folded into list_capabilities) and no related/recommendation or category-tree browsing beyond best_sellers departments, but core workflows are covered.