European IP Data & Analytics — IPRATE

eu.ipratePublished by iprate.euOfficial publisherv0.4.0Updated Oct 25, 2026

Free anonymous selections from released European IP profiles, rankings, and market statistics.

VerifiedStreamable HTTPWeb executableKnowledge & MemoryBusiness & Commerce

Overview

AI-generated overview

A free, anonymous remote MCP server that lets an assistant look up released European IP firm and attorney profiles, rankings and market statistics.

What you can do
Find IP firms or attorneys by name, jurisdiction, right type, Nice class or leading-client match, and read one published profile by ID or slug. Retrieve precomputed market snapshots with leading firms for a cohort, and check which jurisdictions and IP-right questions a release covers. Search open research documents about European IP firms and markets, then fetch one in full with its public citation URL and release metadata.
When to use it
Useful when researching European trademark or patent counsel, comparing firms in a country, or checking what IPRATE publishes about a named firm or market. It suits assistants that need citable, release-stamped IP market data rather than live register queries.
Requirements
A remote streamable HTTP endpoint; no account, key or local runtime is needed for the free path. The README also documents a separate paid Data API endpoint that requires an issued key sent in the X-IPRATE-API-Key header.
Before you install
The free endpoint is read-only and anonymous, with in-memory rate limiting. The README describes a separate paid Data API adapter whose calls are billed and quota-checked, and whose create_watch tool is a persistent write requiring confirmation; that adapter needs the X-IPRATE-API-Key credential. Returned register text is quoted untrusted data.

Installation

In SourceWeft

  1. Open European IP Data & Analytics — IPRATE in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Web executable via Streamable HTTP. Remote servers run from the web runtime once configured in a workspace.

Use with an agent

Paste this prompt into Claude Code, Codex, Cursor or another agent — or into SourceWeft chat. The agent reads this server's install guide, shows you its publisher, endpoint and sign-in, and adds it to your client once you agree. Keys never go into a command.

Read https://sourceweft.com/mcp/eu.iprate%2Fip-analytics/install.md and connect the MCP server it describes to my client. Before changing any config, show me its publisher, endpoint and sign-in method and wait for my OK. Never put an API key in a command or in a file I would commit; ask me where it should come from.

Read the install guide the agent follows

Other MCP clients

Add this to your client's mcpServers config.

{
  "mcpServers": {
    "european-ip-data-analytics-iprate": {
      "type": "http",
      "url": "https://mcp.iprate.eu/mcp"
    }
  }
}

README

IPRATE Free Public MCP

Open-source adapter for IPRATE's anonymous, read-only MCP endpoint:

text
https://mcp.iprate.eu/mcp

The production endpoint is the Cloudflare Worker in worker/. It exposes bounded selections from one completed IPRATE static release through six tools:

  • find_ip_representatives
  • get_ip_representative_profile
  • get_ip_market_snapshot
  • get_iprate_coverage
  • search and fetch, the retrieval pair ChatGPT deep research and company knowledge call (at most ten document references per search, one document per fetch)

Every tool declares an output schema. The Python package in src/ is the self-host reference implementation of the first four tools.

Connect from your assistant

Free path, no account and no key.

Claude Code:

bash
claude mcp add --transport http iprate https://mcp.iprate.eu/mcp

Claude Desktop and claude.ai: Settings, Connectors, add a custom connector with the URL https://mcp.iprate.eu/mcp.

Cursor (~/.cursor/mcp.json):

json
{ "mcpServers": { "iprate": { "url": "https://mcp.iprate.eu/mcp" } } }

ChatGPT: a developer-mode custom connector for one user, or a workspace connector added by a Business or Enterprise admin. Deep research uses the search and fetch tools.

OpenAI Responses API:

json
{ "type": "mcp", "server_label": "iprate", "server_url": "https://mcp.iprate.eu/mcp", "require_approval": "never" }

Paid Data API path, with the key issued in your IPRATE account:

bash
claude mcp add --transport http iprate-data https://mcp.iprate.eu/data-api/mcp --header "X-IPRATE-API-Key: YOUR_KEY"

Prices and the order flow are on https://iprate.eu/developers/data-api/. The setup page with the same steps is https://iprate.eu/developers/mcp/.

The same Worker also contains a separate paid Data API adapter at:

text
https://mcp.iprate.eu/data-api/mcp

It exposes the eight Data API v1 tools: get_coverage, resolve_entity, get_portfolio, get_trademark, search_trademarks, find_representative, get_firm_profile, and create_watch. Custom MCP clients send an issued API credential as X-IPRATE-API-Key. This is an explicit API-key adapter contract; it does not accept or forward OAuth bearer tokens. The backend rechecks scopes, office and field entitlements, quotas, idempotency, and billing for every tool call. create_watch is marked as a persistent write requiring client confirmation. Returned register text is quoted untrusted data.

When the DATA_API_WORKER_SECRET secret binding is configured (wrangler secret put DATA_API_WORKER_SECRET, the same value as the API's DATA_API_WORKER_SECRET), every relayed request carries the end client's address in X-IPRATE-Client-IP with X-IPRATE-Worker-Timestamp and an HMAC-SHA256 X-IPRATE-Worker-Signature over <timestamp>.<address>, so the API's pre-authentication limiter buckets by end client instead of the Worker's egress address. Without the secret the headers are omitted and the API applies its shared relay bucket.

It has no database, ORM, SQL, database credential, application-API fallback, or private IPRATE package dependency. Ratings and statistics are copied from released JSON assets; they are never recomputed by the MCP service. Anonymous calls are rate limited in memory (HTTP 429 with Retry-After; results are never silently truncated), and GET /healthz reports whether one internally consistent release is servable.

Static input

Set IPRATE_MCP_ASSET_ROOT to a completed /data/v1-shaped release directory containing manifest.json, the released cohort and global assets, and mcp/v0.1/catalog.json. Set IPRATE_MCP_ASSET_BASE_URL to the public URL used in source citations.

The catalogue is derived only from static profile assets:

bash
iprate-free-mcp-build-catalog /path/to/data/v1

The builder fails if required indexes are missing, checksums differ, profile counts do not match the release manifest, or a profile asset is unreadable. It never imports or connects to IPRATE's database, API, models, or pipeline.

Release snapshots

In production the adapter never reads the live, mutating export tree. A refresher (iprate-free-mcp-refresh-release) watches the live manifest.json and, when a new release_id appears, copies every manifest-declared asset into an immutable per-release snapshot (checksum-verified against the manifest), builds the catalogue inside that snapshot, and atomically repoints a current symlink. The server keeps serving the previous complete release until the next one is activated; a release that changes while it is being snapshotted is abandoned and retried on the next pass. Snapshots contain no entity profile files — profile data enters only through the derived catalogue — and the catalogue is never written into the publicly served tree.

Development

bash
python -m venv .venv.venv/bin/pip install -e ".[dev]".venv/bin/pytest.venv/bin/ruff check .

Run the adapter against a compatible static release:

bash
export IPRATE_MCP_ASSET_ROOT=/path/to/data/v1export IPRATE_MCP_ASSET_BASE_URL=https://iprate.eu/data/v1iprate-free-mcp

The local endpoint is http://127.0.0.1:8000/mcp; liveness is GET /healthz.

Production containers

compose.yaml runs two hardened services built from the same image:

  • mcp — the server, unprivileged and read-only, mounting only the snapshot directory (read-only) and serving IPRATE_MCP_ASSET_ROOT=/data/current. It joins the existing runtime_default network for Cloudflare Tunnel routing and publishes no host port.
  • refresher — the snapshot builder, with network_mode: none, mounting the live export tree read-only and the snapshot directory read-write.
bash
docker compose up -d --build

Environment overrides: IPRATE_STATIC_ROOT (live export tree), IPRATE_MCP_RELEASES_ROOT (snapshot directory), IPRATE_MCP_REFRESH_UID/GID (owner of the snapshot directory), IPRATE_MCP_RATE_LIMIT_PER_MINUTE (default 120). Neither service receives a database or application secret.

Documentation: https://iprate.eu/developers/mcp/

Cloudflare Worker (production)

The Worker serves POST /mcp, GET /healthz, and the OpenAI domain-verification token at GET /.well-known/openai-apps-challenge. It reads one immutable release from the private R2 bucket that the release builder (iprate-free-mcp-build-worker-release) fills.

bash
cd workernpm cinpx tsc --noEmitnpx vitest run

Response contract since 0.4.0: responses state the release through release_id and as_of. They no longer carry release incident records, the release status word, or a release-age notice. hold_state still reports held or partial.

The domain-verification token is a Worker variable. When the OpenAI plugin portal issues it, add OPENAI_APPS_CHALLENGE = "<token>" under [vars] in worker/wrangler.toml before deploying; wrangler deploy replaces variables set only in the dashboard. Without the variable the path answers 404.

ChatGPT plugin package

plugin/ is the package uploaded to the OpenAI plugin portal (Agent Plugins format): plugin.json with listing, review cases and release notes under extensions.com.openai, mcp.json pointing at the production endpoint, the select-european-ip-counsel skill, and the logo. tests/test_plugin_package.py checks the portal's field limits and that the review cases name the Worker's actual tools.

Build the upload from the folder contents, not the folder itself:

bash
python -c "import pathlib, zipfile; root = pathlib.Path('plugin'); z = zipfile.ZipFile('dist/iprate-plugin.zip', 'w', zipfile.ZIP_DEFLATED); [z.write(p, p.relative_to(root).as_posix()) for p in sorted(root.rglob('*')) if p.is_file()]; z.close()"

The video walkthrough URL (review.demo_recording_url) and any country allowlist (publication.countries) are added before submission. Reviewer credentials are not needed: the server is anonymous.

Licence

Adapter code is Apache-2.0. IPRATE data and website content have separate terms and licences; this code licence does not grant rights to them or to IPRATE trademarks.

Source: README.md at commit 7609fff

Tools

6

find_ip_representatives

Find published IP representatives

Find up to five firms or attorneys from the selected static release. Use this for a name, jurisdiction, right type, released leading Nice class, or released leading-client match. At least one filter is required.

Parametersnametierlimitwindowright_typeclient_namejurisdictionnice_classesrepresentative_type

get_ip_representative_profile

Get a public representative evidence profile

Return the bounded public evidence profile for one published IP firm or attorney. Provide exactly one numeric representative_id or public profile slug.

Parametersslugrepresentative_idrepresentative_type

get_ip_market_snapshot

Get a European IP market snapshot

Return one already-computed static cohort snapshot and up to five leading released firms. No dates, raw records, SQL, arbitrary grouping, or free-form question are accepted.

Parameterstierrequiredwindowrequiredright_typerequiredjurisdictionrequired

get_iprate_coverage

Check IPRATE data coverage

Explain which static release cohorts support a jurisdiction and IP-right question. Returns held, partial, or not-covered state, release assets, counts, and supported fields.

Parameterstierright_typejurisdiction

fetch

Fetch one IPRATE evidence document

Use this after search to read one document in full: a published representative profile with its released ratings, a market snapshot with released statistics and leading firms, or a coverage summary. Returns the text, the public URL to cite, and release metadata.

Parametersidrequired

Version history

2
  1. v0.4.0LatestOct 3, 2026
  2. v0.3.0Sep 16, 2026

Score

100% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 daysRequired25 pts
  • Auth declaredRequired10 pts
  • Tools listedRequired15 pts
  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts