eu-reg-mcp 0.1.0__tar.gz

eu_reg_mcp-0.1.0/.github/workflows/ci.yml

@@ -0,0 +1,15 @@
+name: ci
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - run: uv sync --extra dev
+      - run: uv run pytest -q

eu_reg_mcp-0.1.0/.gitignore

@@ -0,0 +1,6 @@
+__pycache__/
+*.pyc
+.pytest_cache/
+.venv/
+.DS_Store
+dist/

eu_reg_mcp-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Sebastian Förste
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

eu_reg_mcp-0.1.0/Makefile

@@ -0,0 +1,10 @@
+.PHONY: install test run
+
+install:
+	uv sync --extra dev
+
+test:
+	uv run pytest -q
+
+run:
+	uv run eu-reg-mcp

eu_reg_mcp-0.1.0/PKG-INFO

@@ -0,0 +1,124 @@
+Metadata-Version: 2.5
+Name: eu-reg-mcp
+Version: 0.1.0
+Summary: MCP server giving AI agents cited, review-gated grounding in EU regulation: ESMA MiCAR register search, register change tracking, white paper linting, EU AI Act classification.
+Project-URL: Homepage, https://github.com/sebastianfoerste/eu-reg-mcp
+Project-URL: Repository, https://github.com/sebastianfoerste/eu-reg-mcp
+Project-URL: Issues, https://github.com/sebastianfoerste/eu-reg-mcp/issues
+Author: Sebastian Förste
+License-Expression: MIT
+License-File: LICENSE
+Keywords: ai-act,compliance,crypto,esma,eu,mcp,micar,model-context-protocol,regulation
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Software Development :: Libraries
+Requires-Python: >=3.13
+Requires-Dist: mcp<2,>=1.2
+Provides-Extra: classify
+Requires-Dist: eu-ai-act-classifier>=0.2.0; extra == 'classify'
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Provides-Extra: lint
+Requires-Dist: micar-whitepaper-linter>=0.2.0; extra == 'lint'
+Description-Content-Type: text/markdown
+
+<!-- mcp-name: io.github.sebastianfoerste/eu-reg-mcp -->
+# eu-reg-mcp
+
+[![CI](https://github.com/sebastianfoerste/eu-reg-mcp/actions/workflows/ci.yml/badge.svg)](https://github.com/sebastianfoerste/eu-reg-mcp/actions/workflows/ci.yml)
+
+**Let your AI agent check the ESMA MiCAR register and cite EU AI Act provisions instead of guessing — with every answer framed as a fact or a candidate flag for human review.**
+
+Scope today: the ESMA interim MiCAR register (search, summary, change tracking), MiCAR white paper linting, and EU AI Act risk-tier classification. Other EU regulation is not covered.
+
+## Try it
+
+```bash
+claude mcp add eu-reg -- uvx eu-reg-mcp
+```
+
+Then ask your agent: *"Is `<entity>` on the MiCAR register, and what format is its white paper?"* The register tools need no API key. Claude Desktop config, the lint/classify extras and other clients: [docs/tools.md](docs/tools.md#data-sources-and-network-use).
+
+<details>
+<summary>Claude Desktop (<code>claude_desktop_config.json</code>)</summary>
+
+```json
+{
+  "mcpServers": {
+    "eu-reg": {
+      "command": "uvx",
+      "args": ["eu-reg-mcp"]
+    }
+  }
+}
+```
+</details>
+
+## Example output
+
+> **Synthetic data.** Produced on 2026-10-14 by running this repo's register code (`register.search_entries` / `register.recent_changes`) against the **synthetic test fixtures** in `tests/fixtures/`, not the live register. The entity and its LEI are test data. Output shape is what a live call returns; values are not.
+
+`search_micar_register(query="skygate")`:
+
+```json
+{
+  "snapshot_date": "2026-07-07",
+  "total_matches": 1,
+  "results": [
+    {
+      "register": "other-wp",
+      "entity_name": "SKYGATE Network GmbH",
+      "lei": "984500BBFE52FE449926",
+      "member_state": "AT",
+      "authority": "FMA",
+      "whitepaper_url": "WWW.SKYGATETOKEN.AT",
+      "format_class": "unspecified",
+      "register_last_update": "30/12/2024"
+    }
+  ],
+  "truncated": false,
+  "notice": "Facts from the public ESMA interim MiCAR register (Art. 109 MiCAR) as mirrored by the MiCAR Register Observatory. No legal assessment of any entity; not legal advice."
+}
+```
+
+`micar_register_changes(change_type="removed")` returns the same kind of cited record (snapshot date, register, entity) and notes that *"the register itself does not state why"* an entry was removed. The lint and classify tools return pinpoint citations (MiCAR Annex I–III; AI Act Art. 5 / Annex III / GPAI); I have not run them for this README, so no sample is shown. [UNVERIFIED: live output of those two tools]
+
+## Tools
+
+| Tool | What it does | Needs |
+| --- | --- | --- |
+| `search_micar_register` | Look up an entity by name or LEI in the ESMA interim MiCAR register | — |
+| `micar_register_summary` | Register totals and white paper format coverage | — |
+| `micar_register_changes` | New filings, changed entries, removals | — |
+| `lint_micar_whitepaper` | Deterministic Annex I–III lint of a draft white paper | `lint` extra |
+| `classify_eu_ai_act_system` | AI Act risk tier with pinpoint citations | `classify` extra |
+
+Parameters, data sources, caching and network requirements: [docs/tools.md](docs/tools.md).
+
+## Why not just ask an LLM?
+
+An ungrounded model can invent register entries or provisions. Here, register answers come from a dated snapshot of the public register, and lint/classification output comes from deterministic rules with citations. **Review-gated** means the output is a candidate flag pending human legal review — never a legal assessment or legal advice. Details: [docs/guardrails.md](docs/guardrails.md).
+
+## Links
+
+- Register data: [micar-register-observatory](https://github.com/sebastianfoerste/micar-register-observatory) (weekly snapshots of the public ESMA register, Art. 109 VO (EU) 2023/1114)
+- Linter: [micar-whitepaper-linter](https://github.com/sebastianfoerste/micar-whitepaper-linter) (36 rules mapped to Annex I–III) · Classifier: [eu-ai-act-classifier](https://github.com/sebastianfoerste/eu-ai-act-classifier) — both MIT-licensed
+- Development and tests: [docs/development.md](docs/development.md)
+
+## Part of the EU regulation as code suite
+
+By [sebastianfoerste](https://github.com/sebastianfoerste):
+
+- [awesome-eu-fintech-regulation](https://github.com/sebastianfoerste/awesome-eu-fintech-regulation)
+- [contract-review-eval-harness](https://github.com/sebastianfoerste/contract-review-eval-harness)
+- [legal-function-operating-system](https://github.com/sebastianfoerste/legal-function-operating-system)
+- [eu-ai-act-classifier](https://github.com/sebastianfoerste/eu-ai-act-classifier)
+- [micar-whitepaper-linter](https://github.com/sebastianfoerste/micar-whitepaper-linter)
+- **eu-reg-mcp** (this repo)
+- [micar-register-observatory](https://github.com/sebastianfoerste/micar-register-observatory)
+
+## Caveats
+
+Register results are public-register facts with extraction artifacts possible; lint and classification outputs are candidate flags pending human legal review. Nothing here is legal advice. MIT licensed.

eu_reg_mcp-0.1.0/README.md

@@ -0,0 +1,98 @@
+<!-- mcp-name: io.github.sebastianfoerste/eu-reg-mcp -->
+# eu-reg-mcp
+
+[![CI](https://github.com/sebastianfoerste/eu-reg-mcp/actions/workflows/ci.yml/badge.svg)](https://github.com/sebastianfoerste/eu-reg-mcp/actions/workflows/ci.yml)
+
+**Let your AI agent check the ESMA MiCAR register and cite EU AI Act provisions instead of guessing — with every answer framed as a fact or a candidate flag for human review.**
+
+Scope today: the ESMA interim MiCAR register (search, summary, change tracking), MiCAR white paper linting, and EU AI Act risk-tier classification. Other EU regulation is not covered.
+
+## Try it
+
+```bash
+claude mcp add eu-reg -- uvx eu-reg-mcp
+```
+
+Then ask your agent: *"Is `<entity>` on the MiCAR register, and what format is its white paper?"* The register tools need no API key. Claude Desktop config, the lint/classify extras and other clients: [docs/tools.md](docs/tools.md#data-sources-and-network-use).
+
+<details>
+<summary>Claude Desktop (<code>claude_desktop_config.json</code>)</summary>
+
+```json
+{
+  "mcpServers": {
+    "eu-reg": {
+      "command": "uvx",
+      "args": ["eu-reg-mcp"]
+    }
+  }
+}
+```
+</details>
+
+## Example output
+
+> **Synthetic data.** Produced on 2026-10-14 by running this repo's register code (`register.search_entries` / `register.recent_changes`) against the **synthetic test fixtures** in `tests/fixtures/`, not the live register. The entity and its LEI are test data. Output shape is what a live call returns; values are not.
+
+`search_micar_register(query="skygate")`:
+
+```json
+{
+  "snapshot_date": "2026-07-07",
+  "total_matches": 1,
+  "results": [
+    {
+      "register": "other-wp",
+      "entity_name": "SKYGATE Network GmbH",
+      "lei": "984500BBFE52FE449926",
+      "member_state": "AT",
+      "authority": "FMA",
+      "whitepaper_url": "WWW.SKYGATETOKEN.AT",
+      "format_class": "unspecified",
+      "register_last_update": "30/12/2024"
+    }
+  ],
+  "truncated": false,
+  "notice": "Facts from the public ESMA interim MiCAR register (Art. 109 MiCAR) as mirrored by the MiCAR Register Observatory. No legal assessment of any entity; not legal advice."
+}
+```
+
+`micar_register_changes(change_type="removed")` returns the same kind of cited record (snapshot date, register, entity) and notes that *"the register itself does not state why"* an entry was removed. The lint and classify tools return pinpoint citations (MiCAR Annex I–III; AI Act Art. 5 / Annex III / GPAI); I have not run them for this README, so no sample is shown. [UNVERIFIED: live output of those two tools]
+
+## Tools
+
+| Tool | What it does | Needs |
+| --- | --- | --- |
+| `search_micar_register` | Look up an entity by name or LEI in the ESMA interim MiCAR register | — |
+| `micar_register_summary` | Register totals and white paper format coverage | — |
+| `micar_register_changes` | New filings, changed entries, removals | — |
+| `lint_micar_whitepaper` | Deterministic Annex I–III lint of a draft white paper | `lint` extra |
+| `classify_eu_ai_act_system` | AI Act risk tier with pinpoint citations | `classify` extra |
+
+Parameters, data sources, caching and network requirements: [docs/tools.md](docs/tools.md).
+
+## Why not just ask an LLM?
+
+An ungrounded model can invent register entries or provisions. Here, register answers come from a dated snapshot of the public register, and lint/classification output comes from deterministic rules with citations. **Review-gated** means the output is a candidate flag pending human legal review — never a legal assessment or legal advice. Details: [docs/guardrails.md](docs/guardrails.md).
+
+## Links
+
+- Register data: [micar-register-observatory](https://github.com/sebastianfoerste/micar-register-observatory) (weekly snapshots of the public ESMA register, Art. 109 VO (EU) 2023/1114)
+- Linter: [micar-whitepaper-linter](https://github.com/sebastianfoerste/micar-whitepaper-linter) (36 rules mapped to Annex I–III) · Classifier: [eu-ai-act-classifier](https://github.com/sebastianfoerste/eu-ai-act-classifier) — both MIT-licensed
+- Development and tests: [docs/development.md](docs/development.md)
+
+## Part of the EU regulation as code suite
+
+By [sebastianfoerste](https://github.com/sebastianfoerste):
+
+- [awesome-eu-fintech-regulation](https://github.com/sebastianfoerste/awesome-eu-fintech-regulation)
+- [contract-review-eval-harness](https://github.com/sebastianfoerste/contract-review-eval-harness)
+- [legal-function-operating-system](https://github.com/sebastianfoerste/legal-function-operating-system)
+- [eu-ai-act-classifier](https://github.com/sebastianfoerste/eu-ai-act-classifier)
+- [micar-whitepaper-linter](https://github.com/sebastianfoerste/micar-whitepaper-linter)
+- **eu-reg-mcp** (this repo)
+- [micar-register-observatory](https://github.com/sebastianfoerste/micar-register-observatory)
+
+## Caveats
+
+Register results are public-register facts with extraction artifacts possible; lint and classification outputs are candidate flags pending human legal review. Nothing here is legal advice. MIT licensed.

eu_reg_mcp-0.1.0/docs/development.md

@@ -0,0 +1,9 @@
+# Development
+
+```bash
+git clone https://github.com/sebastianfoerste/eu-reg-mcp
+cd eu-reg-mcp
+make install && make test   # `make run` starts the server over stdio
+```
+
+Tests run offline against the fixtures in `tests/fixtures/`. These fixtures are small **synthetic** snapshots, not real register data. The server itself is MIT-licensed.

eu_reg_mcp-0.1.0/docs/guardrails.md

@@ -0,0 +1,8 @@
+# Guardrails: what "review-gated" means here
+
+Every response carries its framing:
+
+- **Register results** are facts from a public register, returned with the snapshot date. A `removed` change record means the entry left the register export; the register does not state why.
+- **Lint and classification outputs** are candidate flags from deterministic rules, **pending human legal review**. Extraction artifacts exist.
+
+Nothing this server returns is a legal assessment of any entity or legal advice — that is a design contract, not a disclaimer. The text lives in the server instructions and in the `notice` field of each response (see `src/eu_reg_mcp/server.py`, `register.py`, `integrations.py`).

eu_reg_mcp-0.1.0/docs/tools.md

@@ -0,0 +1,28 @@
+# Tools reference
+
+| Tool | What it does | Needs |
+| --- | --- | --- |
+| `search_micar_register` | Look up an entity by name or LEI in the ESMA interim MiCAR register: white papers (Titles II–IV), authorised CASPs, non-compliant entities | — |
+| `micar_register_summary` | Current register totals and white paper format coverage | — |
+| `micar_register_changes` | Register movement: new filings, changed entries, removals (withdrawal tracking) | — |
+| `lint_micar_whitepaper` | Deterministic Annex I–III lint of a draft white paper: cited candidate findings | `lint` extra |
+| `classify_eu_ai_act_system` | EU AI Act risk-tier classification with pinpoint Art. 5 / Annex III / GPAI citations | `classify` extra |
+
+## Parameters
+
+- `search_micar_register(query, register_slug="all", member_state="")` — `query` is a case-insensitive entity-name substring or an exact LEI; `register_slug` is one of `all | other-wp | emt-wp | art-wp | casps | ncasp`; `member_state` is an optional two-letter code (DE, IE, MT, …).
+- `micar_register_summary()` — no parameters.
+- `micar_register_changes(limit=50, change_type="all")` — `change_type` is `all | added | changed | removed`.
+- `lint_micar_whitepaper(whitepaper_json)` — a draft white paper as a JSON string.
+- `classify_eu_ai_act_system(profile_json="")` — a SystemProfile as JSON; an empty string returns the schema.
+
+## Data sources and network use
+
+Register data comes from the [MiCAR Register Observatory](https://github.com/sebastianfoerste/micar-register-observatory), which snapshots the public ESMA register weekly (Art. 109 VO (EU) 2023/1114). The server reads the observatory's `data/latest.json` and `data/changelog.jsonl` from GitHub raw (`src/eu_reg_mcp/data.py`) and caches them in memory for one hour. The three register tools therefore need network access to GitHub; they do not contact ESMA directly. If the observatory files are unreachable, the register tools raise an error (no fallback is implemented).
+
+The linter is [micar-whitepaper-linter](https://github.com/sebastianfoerste/micar-whitepaper-linter) (36 rules mapped to Annex I–III); the classifier is [eu-ai-act-classifier](https://github.com/sebastianfoerste/eu-ai-act-classifier). Both are separate packages from their own repositories, MIT-licensed like this server. Without the extras, both tools return an install hint instead of failing.
+
+```bash
+uv pip install 'eu-reg-mcp[lint]'
+uv pip install 'eu-reg-mcp[classify]'
+```

eu_reg_mcp-0.1.0/pyproject.toml

@@ -0,0 +1,48 @@
+[project]
+name = "eu-reg-mcp"
+version = "0.1.0"
+description = "MCP server giving AI agents cited, review-gated grounding in EU regulation: ESMA MiCAR register search, register change tracking, white paper linting, EU AI Act classification."
+readme = "README.md"
+requires-python = ">=3.13"
+license = "MIT"
+license-files = ["LICENSE"]
+authors = [{ name = "Sebastian Förste" }]
+keywords = ["mcp", "model-context-protocol", "eu", "regulation", "micar", "esma", "ai-act", "compliance", "crypto"]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Developers",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.13",
+    "Topic :: Software Development :: Libraries",
+]
+dependencies = ["mcp>=1.2,<2"]
+
+[project.optional-dependencies]
+dev = ["pytest>=8.0"]
+lint = [
+    "micar-whitepaper-linter>=0.2.0",
+]
+classify = [
+    "eu-ai-act-classifier>=0.2.0",
+]
+
+[project.urls]
+Homepage = "https://github.com/sebastianfoerste/eu-reg-mcp"
+Repository = "https://github.com/sebastianfoerste/eu-reg-mcp"
+Issues = "https://github.com/sebastianfoerste/eu-reg-mcp/issues"
+
+[project.scripts]
+eu-reg-mcp = "eu_reg_mcp.server:main"
+
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[tool.hatch.metadata]
+allow-direct-references = true
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/eu_reg_mcp"]
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]

eu_reg_mcp-0.1.0/server.json

@@ -0,0 +1,21 @@
+{
+  "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
+  "name": "io.github.sebastianfoerste/eu-reg-mcp",
+  "title": "EU Regulatory Grounding",
+  "description": "Cited, review-gated grounding in EU regulation: ESMA MiCAR register search and change tracking, white paper linting, EU AI Act classification.",
+  "repository": {
+    "url": "https://github.com/sebastianfoerste/eu-reg-mcp",
+    "source": "github"
+  },
+  "version": "0.1.0",
+  "packages": [
+    {
+      "registryType": "pypi",
+      "registryBaseUrl": "https://pypi.org",
+      "identifier": "eu-reg-mcp",
+      "version": "0.1.0",
+      "runtimeHint": "uvx",
+      "transport": { "type": "stdio" }
+    }
+  ]
+}

eu_reg_mcp-0.1.0/src/eu_reg_mcp/data.py

@@ -0,0 +1,60 @@
+"""Fetch and cache the observatory's published register data.
+
+The MiCAR Register Observatory commits its normalized snapshot and changelog
+to GitHub; this module reads those raw files so every MCP client sees the
+same weekly, review-gated dataset without touching ESMA directly.
+"""
+
+from __future__ import annotations
+
+import json
+import time
+import urllib.request
+from typing import Any
+
+OBSERVATORY_RAW = (
+    "https://raw.githubusercontent.com/sebastianfoerste/micar-register-observatory/main"
+)
+LATEST_URL = f"{OBSERVATORY_RAW}/data/latest.json"
+CHANGELOG_URL = f"{OBSERVATORY_RAW}/data/changelog.jsonl"
+
+USER_AGENT = "eu-reg-mcp/0.1 (+https://github.com/sebastianfoerste/eu-reg-mcp)"
+
+CACHE_TTL_SECONDS = 3600  # register moves weekly; an hour is generous
+
+_cache: dict[str, tuple[float, Any]] = {}
+
+
+def _get(url: str) -> bytes:
+    request = urllib.request.Request(url, headers={"User-Agent": USER_AGENT})
+    with urllib.request.urlopen(request, timeout=30) as response:
+        return response.read()
+
+
+def _cached(key: str, url: str, parse: Any) -> Any:
+    now = time.monotonic()
+    hit = _cache.get(key)
+    if hit is not None and now - hit[0] < CACHE_TTL_SECONDS:
+        return hit[1]
+    value = parse(_get(url))
+    _cache[key] = (now, value)
+    return value
+
+
+def load_snapshot() -> dict[str, Any]:
+    return _cached("latest", LATEST_URL, lambda raw: json.loads(raw.decode("utf-8")))
+
+
+def load_changelog() -> list[dict[str, Any]]:
+    def parse(raw: bytes) -> list[dict[str, Any]]:
+        return [
+            json.loads(line)
+            for line in raw.decode("utf-8").splitlines()
+            if line.strip()
+        ]
+
+    return _cached("changelog", CHANGELOG_URL, parse)
+
+
+def clear_cache() -> None:
+    _cache.clear()

eu_reg_mcp-0.1.0/src/eu_reg_mcp/integrations.py

@@ -0,0 +1,95 @@
+"""Optional integrations with the linter and classifier packages.
+
+Both packages are MIT-licensed and live in their own repositories. They
+install via the `lint` and `classify` extras and degrade to an install hint
+when absent. Findings from either tool are candidate flags requiring human
+legal review, never confirmed deficiencies or legal conclusions — that
+framing is part of the contract.
+"""
+
+from __future__ import annotations
+
+import json
+import shutil
+import subprocess
+import tempfile
+from pathlib import Path
+from typing import Any
+
+REVIEW_NOTICE = (
+    "Deterministic rule output. Every finding is a candidate gap in the "
+    "submitted text pending human legal review; extraction artifacts occur. "
+    "Not legal advice."
+)
+
+
+def lint_whitepaper(whitepaper_json: str) -> dict[str, Any]:
+    binary = shutil.which("micar-lint")
+    if binary is None:
+        return {
+            "available": False,
+            "hint": (
+                "micar-lint is not installed. Install the optional extra: "
+                "`uv pip install 'eu-reg-mcp[lint]'` "
+                "(see github.com/sebastianfoerste/micar-whitepaper-linter)."
+            ),
+        }
+    try:
+        payload = json.loads(whitepaper_json)
+    except json.JSONDecodeError as error:
+        return {"available": True, "error": f"whitepaper_json is not valid JSON: {error}"}
+
+    with tempfile.NamedTemporaryFile(
+        "w", suffix=".json", delete=False, encoding="utf-8"
+    ) as handle:
+        json.dump(payload, handle)
+        path = handle.name
+    try:
+        result = subprocess.run(
+            [binary, path, "--json"],
+            capture_output=True,
+            text=True,
+            timeout=120,
+            check=False,
+        )
+    finally:
+        Path(path).unlink(missing_ok=True)
+    if result.stdout.strip():
+        try:
+            report = json.loads(result.stdout)
+        except json.JSONDecodeError:
+            report = {"raw_output": result.stdout[:5000]}
+        return {"available": True, "report": report, "notice": REVIEW_NOTICE}
+    return {
+        "available": True,
+        "error": (result.stderr or "linter produced no output")[:2000],
+    }
+
+
+def classify_ai_system(profile_json: str) -> dict[str, Any]:
+    try:
+        from eu_ai_act_classifier.local_api import classify_payload, schema_payload
+    except ImportError:
+        return {
+            "available": False,
+            "hint": (
+                "eu-ai-act-classifier is not installed. Install the optional "
+                "extra: `uv pip install 'eu-reg-mcp[classify]'` "
+                "(see github.com/sebastianfoerste/eu-ai-act-classifier)."
+            ),
+        }
+    if not profile_json.strip():
+        return {
+            "available": True,
+            "profile_schema": schema_payload(),
+            "notice": "Empty profile: returning the expected SystemProfile schema.",
+        }
+    try:
+        payload = json.loads(profile_json)
+    except json.JSONDecodeError as error:
+        return {"available": True, "error": f"profile_json is not valid JSON: {error}"}
+    return {
+        "available": True,
+        "classification": classify_payload(payload),
+        "notice": REVIEW_NOTICE,
+    }

eu_reg_mcp-0.1.0/src/eu_reg_mcp/register.py

@@ -0,0 +1,105 @@
+"""Pure query functions over the observatory snapshot."""
+
+from __future__ import annotations
+
+from collections import Counter
+from typing import Any
+
+MAX_RESULTS = 25
+
+NOTICE = (
+    "Facts from the public ESMA interim MiCAR register (Art. 109 MiCAR) as "
+    "mirrored by the MiCAR Register Observatory. No legal assessment of any "
+    "entity; not legal advice."
+)
+
+
+def _all_entries(snapshot: dict[str, Any]) -> list[dict[str, Any]]:
+    return [
+        entry
+        for register in snapshot.get("registers", [])
+        if register.get("fetched")
+        for entry in register.get("entries", [])
+    ]
+
+
+def search_entries(
+    snapshot: dict[str, Any],
+    query: str,
+    register: str = "all",
+    member_state: str = "",
+) -> dict[str, Any]:
+    needle = query.strip().lower()
+    state = member_state.strip().upper()
+    matches = []
+    for entry in _all_entries(snapshot):
+        if register != "all" and entry.get("register_slug") != register:
+            continue
+        if state and entry.get("member_state", "").upper() != state:
+            continue
+        name = entry.get("entity_name", "").lower()
+        lei = entry.get("lei", "").lower()
+        if needle and needle not in name and needle != lei:
+            continue
+        matches.append(
+            {
+                "register": entry.get("register_slug"),
+                "entity_name": entry.get("entity_name"),
+                "lei": entry.get("lei"),
+                "member_state": entry.get("member_state"),
+                "authority": entry.get("authority"),
+                "whitepaper_url": entry.get("wp_url"),
+                "format_class": entry.get("format_class"),
+                "register_last_update": entry.get("last_update"),
+            }
+        )
+    return {
+        "snapshot_date": snapshot.get("snapshot_date"),
+        "total_matches": len(matches),
+        "results": matches[:MAX_RESULTS],
+        "truncated": len(matches) > MAX_RESULTS,
+        "notice": NOTICE,
+    }
+
+
+def summarize(snapshot: dict[str, Any]) -> dict[str, Any]:
+    registers = [
+        {
+            "register": register.get("slug"),
+            "title": register.get("title"),
+            "entries": len(register.get("entries", [])),
+            "fetched": register.get("fetched"),
+        }
+        for register in snapshot.get("registers", [])
+    ]
+    formats = Counter(
+        entry.get("format_class")
+        for register in snapshot.get("registers", [])
+        if register.get("kind") == "whitepaper" and register.get("fetched")
+        for entry in register.get("entries", [])
+    )
+    return {
+        "snapshot_date": snapshot.get("snapshot_date"),
+        "registers": registers,
+        "whitepaper_format_coverage": dict(formats),
+        "notice": NOTICE,
+    }
+
+
+def recent_changes(
+    changelog: list[dict[str, Any]], limit: int = 50, change_type: str = "all"
+) -> dict[str, Any]:
+    records = [
+        record
+        for record in changelog
+        if record.get("change") != "baseline"
+        and (change_type == "all" or record.get("change") == change_type)
+    ]
+    records = list(reversed(records))  # newest snapshot entries last in file
+    return {
+        "total_changes": len(records),
+        "changes": records[: max(1, min(limit, 200))],
+        "notice": NOTICE
+        + " A 'removed' record means the entry left the register export; the "
+        "register itself does not state why.",
+    }