OpenShopGraph

Open, verified shop database for AI agents: products, offers, price comparison, trust and coupons.

CommunityOnlineAPI key requiredGlobalFreeRead-only

What it can do

    What data it sees

    Do you need an account

    An API key from the service settings is required

    Open, verified shop database for AI agents: products, offers, price comparison, trust and coupons.

    How to connect

    How to connect OpenShopGraph to Claude

    The server address to paste into Claude settings:

    https://api.openshopgraph.com/mcp

    1. Open Claude (claude.ai or the desktop app).
    2. Go to Settings → Connectors.
    3. Click Add custom connector.
    4. Paste the server address copied below into Remote MCP server URL and click Add.
    5. In a chat, click + → Connectors and switch the new connector on.

    Custom connectors are available on Free, Pro, Max, Team and Enterprise plans (Free is limited to one). On Team and Enterprise an organization Owner adds the connector first under Organization settings → Connectors.

    Authorization

    After you click Add, a sign-in window for the service opens. Sign in with your own account and approve access. Claude never sees your password.

    How to connect OpenShopGraph to ChatGPT

    The server address to paste into ChatGPT settings:

    https://api.openshopgraph.com/mcp

    1. Open ChatGPT in a browser (chatgpt.com). A Plus, Pro, Business, Enterprise or Edu plan is required.
    2. Turn on developer mode once: Settings → Apps → Advanced settings → Developer mode.
    3. Open Settings → Connectors and click Create.
    4. Fill in the form: Name (anything), Description (one line about what the service does), MCP server URL (copy it below).
    5. Under Authentication choose OAuth if the service requires sign-in, otherwise None. Click Create.
    6. In a new chat open + → Apps/Connectors and enable the connector.

    OpenAI has renamed this section before (Connectors → Apps/Plugins). If the label differs, search settings for "developer mode". On Business/Enterprise workspaces an admin must allow custom connectors first.

    Authorization

    On first use ChatGPT opens the service's sign-in window. Sign in and approve access.

    How to connect OpenShopGraph to Cursor

    The server address to paste into Cursor settings:

    https://api.openshopgraph.com/mcp

    Fastest: click Open in Cursor below and confirm the prompt.

    Manually:

    1. In Cursor open Settings → Cursor Settings → MCP and click Add new global MCP server.
    2. Paste the JSON copied below into ~/.cursor/mcp.json (per project: .cursor/mcp.json in the repo root).
    3. Save the file. The server appears in the MCP list; authorize it there if asked.

    Authorization

    If the service needs sign-in, an authorize button appears next to the server in the MCP list.

    Server tool list (8)

    Raw names from tools/list. Only developers need these.

    find_shopSearch merchants/shops by free-text query (name or domain fragment), optionally filtered by country (ISO 3166-1 alpha-2, e.g. "DE") and/or by an EXACT category slug. Returns the public shop view only — shops still under internal review are never returned. `results` carries at most `limit` entries (default 50, max 100); `total_count` is the full match count before that cap and `truncated` is true when more results exist beyond this page — raise `offset` (default 0) to page through them. Multi-word queries are matched conjunctively: every word must be evidenced on a shop (name, domain or category). `ignored_terms` lists the words that matched nothing anywhere in the corpus and were therefore dropped to avoid an empty answer — a non-empty `ignored_terms` means the results answer a NARROWER question than you asked, so do not present them as satisfying those words. `category_truncated` is true when category resolution hit its internal cap and the match set is a subset. `category`, if given, must be one of the slugs returned by `list_categories` — call that tool first, do not guess a slug; an unrecognized slug is rejected with an error (never a silent empty `results`). `category` is additive to `query` (AND, not OR) — pass `query: ""` to browse a category without a name/domain filter. If many shops match, grouping by country or category keeps the list scannable.
    list_categoriesEnumerate the shop category taxonomy: every category slug currently carried by at least one public shop, together with how many public shops carry it. Call this BEFORE passing `category` to find_shop — the taxonomy is flat (no parent/child hierarchy, e.g. "fashion" and "womens-clothing" are siblings, not nested) and its slugs come from observed data, not a fixed enum baked into this tool, so a slug cannot be reliably guessed. `find_shop(category: <slug>)` only accepts a slug returned here; giving it anything else fails with an error instead of silently returning no results. Coverage measured 2026-09-03: 171 slugs across 3036 of 3100 active shops (97.9%) — a shop can carry more than one category, so counts do not sum to the total shop count.
    get_shopFetch a single merchant/shop by id or domain (public view). Returns "shop not found" for shops still under internal review, identical to a genuinely unknown domain — reviewers cannot be distinguished from typos. The offers array is present for schema stability but empty until merchant offers are ingested (Offer table measured empty on 2026-07-30).
    list_couponsList currently USABLE discount codes/coupons, optionally filtered by shop_domain (e.g. "otto.de"). A code is withheld from `results` when it is EXPIRED (valid_until has passed), when a real checkout test PROVED it does not work (verification_status "verified_rejected"), or when a display policy suppresses it on this channel/country. Nothing is withheld silently: `total_count` is how many codes the query found before any of that, and `withheld` breaks the difference down by reason ({expired, verified_rejected, reserved_test_domain, policy}) — total_count minus results.length always equals the sum of `withheld`, and `note` states the same balance in one sentence. So an empty `results` NEVER has to be guessed at: read `total_count` and `note` to tell "this shop has no codes at all" apart from "it has codes and every one of them is proven dead". An UNKNOWN or internal-review shop_domain is a different answer again — it is an error ("shop not found"), not an empty list; and a backend failure is reported as a failure, never as an empty list. Each result carries validUntil: an ISO timestamp when the code has a stated expiry, or null when it is open-ended — null means unlimited, NOT unknown-and-expired. Re-check validUntil against the current time before relaying a code. Each result ALSO carries verificationStatus, the verdict of a real checkout test against this exact code: "verified_applied" means the code was proven to work by an actual checkout attempt with proof on file — safe to recommend; "unverified" means no proof exists either way yet — treat it as unknown, NOT as a negative signal, and do not imply it was tested and failed. ("verified_rejected" is defined as before but no longer appears in `results` — it is counted in `withheld.verified_rejected` instead.) lastVerifiedAt is the ISO timestamp of that most recent test, or null if the code was never tested. A non-null lastVerifiedAt on an "unverified" code means it WAS tested but the test was inconclusive — it is no
    get_shipping_policyWhat a shop STATES about its own shipping: which destinations it names, the delivery time it claims, and any free-shipping threshold — plus the date that statement was last checked and where it was read (policy_page or structured_data). This is a shop-level statement, NOT an offer and NOT a quote: it carries no carrier and no shipping price, because OpenShopGraph does not hold that data. Pass country (ISO 3166-1 alpha-2, e.g. "DE") to get a ships_to verdict derived from the stated destination list — "unknown" is a real answer there and means the statement does not settle it. When nothing has been collected for a shop, the reply says so explicitly (collected:false): that is a gap in the data, not a statement that the shop does not ship. Relaying the checked-at date alongside any figure keeps a stale merchant claim from being read as a current fact.
    get_trustTrust-relevant facts OpenShopGraph itself measured about one merchant/shop (by id or domain): TLS certificate issuer and grade, hosting ISP and country, domain registration date, registrar, and accepted payment methods. Returns INDIVIDUAL signals, never an aggregate: every entry carries its own value, `source` (where OpenShopGraph observed it) and `checked_at` (when). There is deliberately no score, rating, percentage or star value — not even an internally computed one; you are given the raw evidence and judge for yourself. A field with no source or no check timestamp is omitted entirely rather than shipped with a null placeholder. If nothing was measured, the answer is status "no_signals_collected" with an empty list — that means NOTHING IS KNOWN, it is not a score of zero and not a negative verdict. Unknown domains, shops under internal review and measured-but-empty shops are intentionally indistinguishable here. `signal_coverage` describes data completeness, not merchant trustworthiness: `present`/`total` count how many of the (up to 7) signal fields OpenShopGraph was able to deliver for this shop, and `missing` lists the rest, each with a `reason` — `not_collected_yet` (OpenShopGraph has never run that check for this shop) or `pipeline_gap` (the check ran — source and timestamp exist — but produced no usable value, e.g. an IP/ASN lookup that resolves a hosting ISP but not a country). A low `present` count is NOT evidence of a low-quality or risky merchant: it usually just means one of OpenShopGraph's own probes came back empty. Do not treat `signal_coverage` as a rating, and do not rank or filter merchants by it as if it were one. Unlike every other tool on this page, this one requires a valid `x-api-token` header — a call without one fails before reaching this tool (401, no exceptions for anonymous/public callers).
    report_codeRecord that a specific shop_domain + code (coupon) pair is wrong in one of four fixed, documented ways (invalid | mismatch | item_restricted | min_order_missing — see reasons enum; no free text anywhere). This is a write: it appends one new report row and NEVER changes the delivered coupon directly — it never deletes or overwrites data, and it does not itself decide anything. Subject to per-source dedup it only queues a re-check (RecheckQueueEntry); only a separate Pruefurteil process may later change what list_coupons returns. Reply is a status text ('wird gemeldet, Pruefung laeuft'), never a verdict.
    report_issueRecord a shop-level or data-fact issue as one fixed, documented reason (no free text). subject_type "shop": unreachable | no_longer_exists | info_outdated | wrong_category | wrong_name_or_domain. subject_type "fact": shipping_cost | return_period | payment_methods | evidence_broken. Same write semantics as report_code: appends one report row, never changes delivered data directly, only queues a re-check subject to per-source dedup.