@pipeworx/amf-filings

French AMF-regulated company disclosures (résultats, franchissements de seuils, information privilégiée, rachats d'actions, etc.) from Info-financière, the French Officially Appointed Mechanism (OAM) — updated as the AMF publishes, not months behind like the community ESEF index.

Part of Pipeworx — an MCP gateway connecting AI agents to 1861+ live data sources. This is an independent, unofficial integration — not affiliated with, endorsed by, or published by the upstream provider.

Tools

• amf_search_filings(company?, isin?, lei?, date_from?, date_to?, language?, limit?, offset?) — search disclosures by issuer (ISIN preferred, LEI often null, or free-text name) and/or a publication date window. date_to is INCLUSIVE (the last day you want included) — set date_from and date_to to the SAME date for a single-day check. Returns the newest first.

• amf_get_filing(id) — fetch one disclosure by its Info-financière record id (uin_idt_uin), as returned by amf_search_filings.

Related MCP server: BODACC MCP Server

Why this pack exists

esef-filings proxies filings.xbrl.org, a community index whose current-year coverage for French issuers is effectively absent — a named external developer (evaluating us for a weekly equity-analysis pipeline) hit this directly and rejected the pack over it. Measured live 2026-09-23, before this pack existed:

France, ESEF index (filings.xbrl.org):  2023: ~ok   2024: ~ok   2025: near-zero

This pack goes to the primary source instead of the aggregator. Measured freshness, 2026-09-23: the newest record in the dataset was timestamped ~1 minute before the query that found it. A date_from/date_to window of "yesterday to today" reliably returns same-day filings — verified with a live call against the deployed gateway (see workers/gateway/src/tool-examples.json for the exact arguments used).

One call per day, not one call per issuer

company, isin, and lei are all optional. Omit them entirely and pass only date_from/date_to to get every issuer's filings for that window in ONE call (paginate with limit/offset only if a single day's market-wide volume exceeds the 100-row cap) — verified live 2026-10-19. A design partner doing a daily check across ~155 issuers was instead calling once per issuer and exhausting a 50-credit-per-call allowance in ~35 calls (fleet #3300); the date-window call is the cheap way to do that check.

⚠️ Six things a caller must know

1. This is a document index, not an XBRL-facts API. Every row is a disclosure EVENT: a title, an issuer, a publication timestamp, and a link to the underlying PDF (occasionally an ESEF/XBRL package). There is no per-fact extraction — no "give me net income for FY2025" lookup. Whether filings.xbrl.org-shaped structured facts are reachable for these same issuers via this OAM was not verified while building this pack; treat it as a separate, open question, not a promise this pack makes.

2. LEI is frequently null. identificationsociete_iso_cd_lei is populated on many but not all rows. ISIN (identificationsociete_iso_cd_isi) is far more reliable — prefer it as the join key. Do not build an LEI-only pipeline against this pack; a miss on LEI does not mean "no filings."

3. Not scoped to France-domiciled issuers. The feed carries any issuer that discloses through a French disseminator — this includes non-French subsidiaries of French groups (e.g. Gabon-domiciled TotalEnergies EP Gabon). country reflects the issuer's own registered country in the AMF's record, not "France" by default.

4. company matches the LEGAL name, not the trade name — with an automatic fallback. The AMF stores issuers under their registered legal name (e.g. "TELEVISION FRANCAISE 1"), and search() is token-based, not substring, so a trade-name query like company: "TF1" matches zero rows there even though 38 genuinely exist in the same window (verified live 2026-10-19, fleet #3300). amf_search_filings now retries against the disclosure TITLE when the legal-name search comes back empty, and the response carries matched_via: "title" plus company_resolved (the real legal name) when that is what found it — pass isin when you have it to skip the fallback entirely.

5. subtype_of_information is normalised to a real null. Some disseminators (307 in particular, 9,519 rows at last check) send the literal STRING "NULL" rather than an actual null. This pack converts any case-insensitive "null" string to a real null before returning it.

6. date_to is INCLUSIVE of that whole UTC day. A one-day check sets date_from and date_to to the SAME date; this pack handles the upstream's own exclusive bound internally (querying one calendar day past it) and echoes the actual bound sent as upstream_date_to_exclusive. This used to not be true: a same-day window (date_from=date_to="2026-10-16") silently returned zero rows where 61 existed, because the upstream bound was passed straight through as an exclusive one — fixed 2026-10-20 (a paying customer running one-day ESEF checks across every issuer hit it directly). A date_from after date_to now returns empty_reason: "empty_window" with an explanatory message instead of a bare zero.

Bilingual issuers file the same disclosure in French and English as two separate rows (each with its own id) within the same minute — amf_search_filings does not merge these; pass language to pick one edition, or group results client-side on the returned language_pair_key (company + type/subtype + publication minute).

Auth

Keyless. No account, no API key.

Data sources

• https://www.info-financiere.gouv.fr/api/explore/v2.1/catalog/datasets/flux-amf-new-prod/records — Opendatasoft Explore v2.1 records API, flux-amf-new-prod dataset (500,000+ records at time of writing). This is the primary AMF disclosure flux, not a mirror or cache of it — Pipeworx queries it live on every call.

Gotchas

• The historic host www.info-financiere.fr 301-redirects every request to www.info-financiere.gouv.fr — this pack calls the .gouv.fr host directly to save the round trip.

• limit is capped at 100 by the upstream API (-1 <= limit <= 100); a higher value 400s rather than silently clamping. This pack clamps to 100 before the request goes out and reports the original value as limit_clamped_from in the response when it does, so the clamp is never silent.

• ODSQL where clauses on uin_dat_amf (the filing timestamp) genuinely bind — verified live with a same-day-exclusion test — unlike the sibling esef-filings date-filter bug this pack was written to not repeat.

• search() on the company-name field is a loose token match (per ODSQL semantics), not an exact-name filter — a query for one issuer can surface a related entity with a similar name (e.g. a subsidiary). Prefer isin when you have it.

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "amf-filings": {
      "url": "https://gateway.pipeworx.io/amf-filings/mcp"
    }
  }
}

What this endpoint actually serves

tools/list at https://gateway.pipeworx.io/amf-filings/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — ask_pipeworx, discover_tools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.

This is deliberate, not multiplexing by accident. The meta-tools are what let a scoped connection answer a question this pack does not cover — via ask_pipeworx, which routes across the whole catalog — without you adding a second MCP server. There is currently no way to mount a pack endpoint without them; if the extra schemas cost you more context than the routing is worth, connect to the full gateway once rather than to several pack endpoints.

Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/mcp"
    }
  }
}

Both URLs reach the same gateway and the same 1861+ data sources. The only difference is which pack's tools are listed directly; ask_pipeworx reaches all of them from either one.

No MCP client? Call it over HTTP

curl -X POST https://gateway.pipeworx.io/v1/tools/amf_search_filings \
  -H 'Content-Type: application/json' \
  -d '{"isin":"FR0013230950","limit":3}'

No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/amf_search_filings. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.

Run it from npm

@pipeworx/mcp-amf-filings is a local stdio bridge to the Pipeworx gateway: your MCP client starts it with npx, and it forwards every call to the same https://gateway.pipeworx.io/amf-filings/mcp endpoint shown above. It needs network access; there is no offline mode.

{
  "mcpServers": {
    "amf-filings": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-amf-filings"]
    }
  }
}

Without a key it calls anonymously, capped at 50 calls a day. A free account raises that to 200 a day: get a key at https://pipeworx.io/signup and add "env": { "PIPEWORX_API_KEY": "<your key>" } to the server entry. The key is sent as a Bearer token, the same way a hosted connection sends it, so calls are metered and billed exactly as they are there.

It lists only this pack's tools by default. Set PIPEWORX_META_TOOLS=1 to also list ask_pipeworx and the other gateway-wide tools.

Behind an HTTPS proxy: Node's built-in fetch does not read HTTPS_PROXY by default. Newer Node releases honour it when NODE_USE_ENV_PROXY=1 is set; check your Node version's documentation.

Using with ask_pipeworx

Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:

ask_pipeworx({ question: "your question about Amf Filings data" })

The gateway picks the right tool and fills the arguments automatically.

More

• Docs and guides

• pipeworx.io

License

MIT