v0.2.1

First release published to PyPI. Install it with uvx amazon-india-seller-mcp.

Fixed

• The publish workflow never ran for v0.1.0 or v0.2.0. The Release workflow
  creates the GitHub Release using GITHUB_TOKEN, and GitHub blocks events
  raised by one workflow's token from starting another, so release: published
  never fired. Publishing now triggers on the v* tag push itself.
• Added a test asserting the version in pyproject.toml matches
  __version__, so the two cannot drift apart between releases.

Note

Package contents are identical to v0.2.0, which was tagged but never uploaded.

Full Changelog: v0.2.0...v0.2.1

v0.2.0

Changed

• The project is now an installable distribution. Install with
  uvx amazon-india-seller-mcp, uv tool install amazon-india-seller-mcp or
  pip install amazon-india-seller-mcp — no clone, and no absolute paths in the
  Claude Desktop config.
• All code moved into the amazon_india_seller_mcp package. The previous
  top-level config, database, services and tools modules would have
  collided with other distributions in site-packages.
• Added an amazon-india-seller-mcp console script and a
  python -m amazon_india_seller_mcp entry point.

Added

• A trusted-publishing workflow for PyPI (OIDC, no API token stored in the repo)
  that runs the full suite, checks the distribution metadata, and verifies the
  wheel installs and registers all 24 tools in a clean environment before upload.

Compatibility

• python server.py still works. The root server.py is now a shim, so existing
  Claude Desktop configurations that point at it by path need no changes.

Full Changelog: v0.1.0...v0.2.0

v0.1.0

First public release. An MCP server that turns Claude Desktop into a product research
assistant for beginner Amazon India sellers — 24 tools, working with zero API keys.

Added

Core research (8 tools)

• research_product — full opportunity report with a weighted 0–100 score
• analyze_product_demand — demand level, trend, seasonality and a launch decision
• analyze_competition — competition level, review barrier, brand dominance, gaps
• calculate_profitability — referral, closing, fulfilment and GST fees, return
  reserve, margin, ROI, break-even and recommended price
• search_suppliers — sourcing research for Parrys, Chennai, Tamil Nadu and India
• analyze_reviews — complaints clustered by theme with supplier-level fixes
• research_keywords — primary, secondary, long-tail and backend search terms
• generate_listing — title, bullets, description, image brief and compliance checks

Sales and competitor intelligence (6 tools)

• calculate_revenue — revenue from units, BSR curves or purchase badges, as a range
• analyze_competitors — per-competitor units, revenue and market share; flags new
  sellers by review count and who clears 300+ units a month
• analyze_purchase_signals — aggregates Amazon's own "bought in past month" badges
• analyze_review_metrics — the review barrier and which listings are beatable
• analyze_evergreen — evergreen vs seasonal scoring from up to 5 years of interest
• analyze_product_images — gallery coverage and a seven-slot image plan

Planning (2 tools)

• find_product_opportunities — screen and rank up to 15 ideas at once
• plan_product_launch — order quantity, budget split, reorder point, payback, timeline

Amazon Ads (3 tools)

• suggest_ppc_keywords — keywords with match type, bid and campaign placement
• calculate_ppc_bids — break-even ACOS and CPC, bid ladder, per-order economics
• plan_ppc_campaign — three-campaign structure, budget split, weekly routine

Live data (5 tools)

• search_web — DuckDuckGo (free, no key), Brave, Serper, Tavily, Google CSE
• scrape_amazon_search — live search results with purchase badges
• scrape_amazon_product — product page with BSR, weight and sales estimate
• scrape_listing_details — full listing teardown, graded 0–100
• scraper_status — what the live-data layer is configured to do

Free live data sources, no API keys

• Google Trends via pytrends for real India search interest and seasonality
• DuckDuckGo web search
• Public amazon.in pages through a guardrailed scraping layer

Data integrity

• Every meaningful output carries source, data_type, confidence and
  last_updated, where data_type is Live, Verified, Estimated, Historical or Demo
• Unparseable values return null, never 0 or a guess
• Modelled figures return a range and name the method that produced them
• Supplier names, prices and MOQs are never fabricated
• Demo mode is deterministic and always labelled

Security

• SSRF protection: scheme, port and resolved IP validated; every redirect hop
  revalidated; loopback, private, link-local and cloud metadata ranges refused
• Credential redaction filter covering library logging such as httpx request URLs
• Prompt-injection scanning and sanitisation of all scraped and searched content,
  surfaced through a content_safety block
• 8 MB response cap, 5-hop redirect limit, validated config file paths

Scraping guardrails

• Domain allowlist, robots.txt enforcement, per-host crawl delay and page budget
• Bot challenges are detected and stop the run; bypass is deliberately not implemented
• Per-field parse coverage reporting, and selector overrides via configuration

Project

• MIT licence, contribution guide, code of conduct and security policy
• CI on Python 3.11, 3.12 and 3.13, plus a security job
• Documentation: setup guide, scraping guide and a 56-prompt library

Known limitations

• BSR-to-units curves and PPC conversion benchmarks are reasoned approximations, not
  calibrated against real sales data. They are labelled Estimated and return ranges.
• The bundled Amazon fee schedule is approximate. Point AMAZON_FEE_CONFIG_PATH at
  your Seller Central rate card for accurate profit figures.
• GST is applied to Amazon's fees but not to the sale price. Since Amazon India prices
  are GST-inclusive, reported margins are optimistic for GST-registered sellers. This
  is the first thing being fixed in 0.2.0.
• Amazon serves bot challenges intermittently, so direct scraping is opportunistic.
  Google Trends and DuckDuckGo are the dependable free sources.
• SP-API and Product Advertising API providers are routed but not implemented; they
  raise a clear error rather than returning fabricated data.

What's Changed

• feat: add live data sources and 12 advanced seller research tools by @Suriya-Ravichandran in #1
• Feat/live data and advanced research tools by @Suriya-Ravichandran in #2
• feat(security): add SSRF protection, credential redaction and injecti… by @Suriya-Ravichandran in #3

New Contributors

• @Suriya-Ravichandran made their first contribution in #1

Full Changelog: https://github.com/Suriya-Ravichandran/amazon-india-seller-mcp/commits/v0.1.0