mcp server

Wafergraph MCP

Read-only MCP server for wafergraph.com's semiconductor & AI supply-chain data: 30 tools, no auth.

Description as published by the maintainer. Source

• version 1.2.0
• active

active — Most recent push to the repository was 2026-07-31.

What this server can do

30 functions, named and described by the server itself. Parameter names are shown because they say more about what a function does than its name usually does.

analyze_portfolio_exposure(holdings)

Given a list of tickers or company ids, report that basket's aggregate exposure across supply-chain segments and countries, and flag where holdings share the same upstream suppliers (correlated single points of failure). Informational supply-chain analysis over public data, not investment advice. Required: holdings.

compare_companies(ids)

Side-by-side comparison of 2-6 companies on the same fields, plus their shared and unique supply-chain counterparties. Cheaper and more aligned than several get_company calls when the question is comparative. Required: ids.

compare_countries(countries)

Side-by-side comparison of 2-5 countries: aligned rows for company count, segment mix, market-position mix, and priced market cap, plus which segments each country is uniquely present in or dominant in, and which segments they all share. country is the company's HEADQUARTERS country only, not a manufacturing-footprint field. A company headquartered here may fabricate, assemble, or test elsewhere — do not read this data as production geography. Required: countries.

filter_companies(limit, offset, public, country, segment, sort_by, has_ticker, subsegment, market_position, max_market_cap_usd_b, min_market_cap_usd_b)

Structured multi-criteria screen over all 615 companies: exact segment/subsegment/country/market_position/public filters plus a market-cap range, sortable and paginated. Use this instead of search_companies when the question is a precise filter ('leader-position analog companies in Japan under $20B') rather than a free-text match. Unknown segment/subsegment/country values just return zero results rather than erroring — call get_segments or list_subsegments first if you're not sure a value is valid.

find_chokepoints(limit, segment)

Rank supply-chain chokepoints: companies many others depend on, weighted by how concentrated their market position is. A chokepoint here means high downstream dependency plus monopoly/leader position, i.e. few substitutes. Scoring is a transparent heuristic over the public dataset, not a proprietary risk model.

find_common_suppliers(limit, segment, company_ids)

The shared-upstream question over a set of companies: given 2-15 company ids/tickers, or a segment id (uses every company in that segment), rank suppliers by how many of the input companies they documentedly serve (e.g. 'serves 9 of 12'), with each supplier's market position and country. Also reports how many input companies had no documented suppliers at all, since that makes a low overlap number ambiguous.

find_consolidation_hotspots(limit, sort_by)

Ranks taxonomy segments by M&A activity by mapping each deal's parties onto their companies' segments (matching by id, then falling back to case-insensitive name), then aggregating deal count and disclosed value per segment. Deals whose parties cannot be resolved to any dataset company are counted in unmapped_deals rather than dropped, so thinly-covered segments aren't silently underrepresented.

find_deals_by_company(company)

Every M&A deal a company took part in, split by role (as acquirer, as target, or other). Matches by dataset id first, then falls back to case-insensitive name matching — necessary because a deal's target is frequently not itself a company in this dataset and carries a null id (e.g. AMD's acquisition of Xilinx lists acquirer id 'amd' but target id null, name 'Xilinx'). Each matched deal carries a match_method ('id' or 'name') so weaker name-only matches are visible to the caller. Required: company.

find_paths_between(to, from, limit, direction, max_depth)

Every documented supply path between two companies, following supplier->customer edges (e.g. 'how does NVIDIA actually depend on Shin-Etsu'). Searches up to max_depth hops in one or both directions and returns each path as an ordered list of companies, shortest first. Capped for combinatorial safety; absence of a path means undocumented, not disproven — see edge_coverage. Required: from, to.

find_similar_companies(id, limit)

Nearest structural neighbours to one focal company, ranked by a transparent Jaccard-similarity score — not a market or competitive judgment. Use search_companies or resolve_ticker first if you only have a ticker or an approximate name, then pass the resolved id here. Required: id.

find_single_source_dependencies(limit, country, segment)

Screen for (customer, subsegment) pairs where the customer has exactly ONE documented supplier in that subsegment — the highest-value documented-concentration risk screen in the dataset. Optionally scoped to customers in one segment or country. Ranked by the sole supplier's downstream importance (its total documented customer count).

get_company(id)

Full allowed profile for one company (by id or exact name) plus its supplier/customer supply-chain edges. Includes key_products (short list of named products/lines). Fields are deliberately limited to established/trust-checked data (see README field-discipline note). Required: id.

get_country_exposure(segment)

Geographic concentration of the supply chain: which countries host the companies in a given segment (or across all 12 segments), ranked by company count. Answers 'how concentrated in Taiwan is advanced lithography' style questions. Country is recorded for all 615 companies.

get_country_profile(country)

Deep profile of one country's presence in wafergraph's semiconductor & AI supply-chain dataset: company count, segment breakdown, market-position breakdown, top companies by market cap, notable monopoly/leader companies, and inbound/outbound supplier-relationship edge counts across this country's border (computed from the supply-chain graph). country is the company's HEADQUARTERS country only, not a manufacturing-footprint field. A company headquartered here may fabricate, assemble, or test elsewhere — do not read this data as production geography. Required: country.

get_dataset_stats

The honesty tool: what this dataset actually contains and where it is thin. Live-computed per-field coverage for companies and deals, last_verified staleness distribution, supply-chain edge coverage, data source mode, and a plain-words list of known limitations. Call this before treating an absence of a company, deal, or edge as evidence it doesn't exist in the real market.

get_deal(id)

Full record for one M&A deal by id: title, type, value, announced date, status, all parties with their resolved company refs where a dataset id exists (and the raw party name where it does not), summary, sources, and the per-deal confidence flag. Use get_deals or find_deals_by_company to find a deal id first. Required: id.

get_deals(query, segment)

Search wafergraph's semiconductor & AI supply-chain M&A corpus (74 acquisitions/mergers, including notable terminated attempts) by title/summary substring and/or segment. Returns a compact list capped at 30 with a total match count.

get_ma_activity_summary(top_n)

Aggregate view of the full 74-deal M&A corpus: counts by year (from announced date), by deal type, and by status; total and median disclosed value; and the largest deals by value. Value figures are computed only over the subset of deals with a disclosed value_usd and are never extrapolated to cover the undisclosed ones.

get_segment_leaders(segment)

Who runs a given layer of the semiconductor & AI supply chain: the companies at monopoly/leader market position in one taxonomy segment (or all 12 if none given), with country and market cap, plus a count of how many companies sit at each position (monopoly/leader/major/challenger/niche) in that segment. country is the company's HEADQUARTERS country only, not a manufacturing-footprint field. A company headquartered here may fabricate, assemble, or test elsewhere — do not read this data as production geography.

get_segments

The wafergraph taxonomy: 12 top-level supply-chain segments (materials through ai_datacenter) and their subsegments, each with a live company count, plus the market_position enum. Use this to discover valid `segment` values for search_companies/get_deals. Segment definitions are a versioned snapshot (see data.taxonomy_snapshot_date) while company counts are computed live.

get_subsegment(limit, offset, segment, subsegment)

All companies in one segment+subsegment pair, as compact refs sorted by market cap descending, plus a market_position breakdown and a country breakdown computed over the FULL matching set (not just the returned page). Use list_subsegments first if you don't know valid segment/subsegment ids. Required: segment, subsegment.

get_supply_chain(id, depth, direction)

Walk the supplier/customer graph from one focal company, up to 2 tiers up (suppliers), down (customers), or both. Mirrors the chain view on wafergraph.com's Explorer. Returns companies grouped by tier plus the edges between them. Required: id.

get_upstream_concentration(id)

For one focal company: break its suppliers down by headquarters country and by segment, report an HHI concentration index (0 = spread evenly, 1 = fully concentrated in one bucket) for each dimension, and name the single most concentrated one. Always reports supplier_edge_coverage because key_suppliers is only ~58% filled dataset-wide — a company with few listed suppliers here may be under-documented, not genuinely un-dependent. country is the company's HEADQUARTERS country only, not a manufacturing-footprint field. A company headquartered here may fabricate, assemble, or test elsewhere — do not read this data as production geography. Required: id.

list_countries(segment)

Every country in wafergraph's semiconductor & AI supply-chain dataset (29 countries across 615 companies) with company count, which segments are present there (with counts), public/private split, and priced market-cap totals. Sorted by company count descending. Optional segment filter. country is the company's HEADQUARTERS country only, not a manufacturing-footprint field. A company headquartered here may fabricate, assemble, or test elsewhere — do not read this data as production geography.

list_subsegments(segment)

Every subsegment across wafergraph's 12-segment taxonomy, each with its live company count and parent segment id/name, optionally filtered to one segment. Use this (or get_segments) to discover valid `subsegment` values before calling get_subsegment or filter_companies. Segment/subsegment names come from a versioned taxonomy snapshot; company counts are computed live and can include subsegment ids present in the company data but not yet in that snapshot (flagged `in_taxonomy: false`).

rank_by_connectivity(limit, metric, country, segment)

Rank companies by documented supply-chain degree: customer count (downstream reach), supplier count (upstream dependence), or total. CRITICAL: degree measures how well a relationship is DOCUMENTED in this curated dataset, not how critical the company actually is — a well-covered firm can outrank a more essential but obscure one. See the caveat field in every response.

rank_by_market_cap(limit, country, segment, market_position)

Top N companies by market cap, optionally restricted to a segment/country/market_position, with the priced-coverage ratio for that scope attached — about 28% of companies dataset-wide have no market_cap_usd_b on file, so a plain top-N list without the coverage number would look more complete than it is.

resolve_ticker(queries)

Batch-resolve up to 25 strings — tickers, company names, or ids, in any mix — to canonical company refs. Call this FIRST whenever you have raw user input (a ticker list, pasted names) and need valid ids before calling other tools; unresolved entries come back with up to 3 suggested close matches instead of just null. Required: queries.

search_companies(query, country, segment)

Search wafergraph's semiconductor & AI supply-chain company dataset (615 companies across 12 segments) by name/one_liner substring and/or segment and/or country. Returns a compact list capped at 25 with a total match count. Use get_segments first if you don't know valid segment ids.

simulate_disruption(limit, country, segment, company_id)

Remove one company, every company in one country, or every company in one segment from the documented supply graph and report the blast radius: which companies lose a documented supplier, how many alternative suppliers they retain in the same subsegment, and which are left with zero documented alternative (ranked first). This is a documented-edge simulation, not a forecast — see the caveat field.

Last successful function declaration observed on 2026-08-13. Source: https://wafergraph-mcp.jwpalm99.workers.dev/mcp. We list what the server declared; we do not call any of these functions.

Endpoint status observed on 2026-08-13. Source: https://wafergraph-mcp.jwpalm99.workers.dev/mcp.

Signals

These are separate measurements of different things. They are deliberately not combined into one score, because a popularity number that mixes website traffic with saves and stars cannot be checked or acted on.

Where to get it

• Repository — https://github.com/jasonpalmer1/wafergraph-mcp
• Remote endpoint (streamable-http) — https://wafergraph-mcp.jwpalm99.workers.dev/mcp

Related, by what their authors tagged them

• io.github.amplichina/madeforai-supplychain-skill — last commit 2026-07-30, shares supply-chain
  MCP connector for human-reviewed manufacturing tasks through MadeForAI hosted fulfillment.
• Compuute MCP Security Scanner — last commit 2026-07-03, shares supply-chain
  Scan any public GitHub MCP-server repo for security issues. 37 MCP-specific L1 rules, 8 languages.
• io.github.CSOAI-ORG/sbom-cyclonedx-mcp — last commit 2026-06-26, shares supply-chain
  Software Bill of Materials generation + validation in CycloneDX 1.6 and SPDX 2.3 formats. Requir...
• TridentChain Security — last commit 2026-07-29, shares supply-chain
  Local supply-chain CVE scanner via OSV/NVD. Scans deps and IDE extensions. No upload.
• io.github.eliottreich/taskbounty-check — last commit 2026-06-21, shares supply-chain
  Local GitHub Actions/CI maintenance check (action pinning, token perms). Not a full security audit.
• Closeread Verify — last commit 2026-06-11, shares supply-chain
  A verified dependency-audit verdict any AI agent can call over MCP, not a raw scan.
• ai.radixia/blog — last commit 2026-08-05, shares cloudflare-workers
  Search and read the Radixia blog (enterprise AI, cloud, open source) via MCP.
• refd — last commit 2026-07-30, shares cloudflare-workers
  Read AI search visibility, competitors, citations, prompts, and answer evidence from refd.
• ai.smithery/TakoData-tako-mcp — last commit 2026-08-06, shares cloudflare-workers
  Provide real-time data querying and visualization by integrating Tako with your agents. Generate o…
• Scry — archived, last commit 2026-05-25, shares cloudflare-workers
  Free IPv4 lookups against a distributed attacker-observation corpus.

These share tags the maintainers applied themselves, such as supply-chain, cloudflare-workers. Common tags like "mcp" or "ai" are ignored for this: agreeing with six hundred other projects is not a similarity.

This is not a recommendation and not a test result. It is a map of what the authors said their work is about.

How the author describes it

Topics the maintainer set on GitHub: cloudflare-workers, mcp, mcp-server, semiconductors, supply-chain.