generated: '2026-08-09'
method: derived
status: candidate
source: openapi/i6eal-open-ai-data-api-openapi.json
search_result:
  official_hosted_server: none
  probes:
  - url: https://i6eal.de/mcp
    method: GET
    status: 404
  - url: https://i6eal.de/.well-known/oauth-protected-resource
    method: GET
    status: 404
  - url: https://i6eal.de/.well-known/ai-plugin.json
    method: GET
    status: 404
  registries_searched:
  - npm (@modelcontextprotocol and "i6eal") — 0 results
  - PyPI "i6eal" — 404
  - github.com/i6eal (13 public repos) — no MCP server repo
  finding: 'i6eal publishes no MCP server, first- or third-party. What it does publish for agents is a
    real, provider-authored llms.txt (harvested verbatim to llms/) that carries the dataset list, the
    citation format, the licence scoping and — unusually — explicit reading rules about what the data
    does and does not claim. The candidate tool list below is DERIVED from the OpenAPI and is API Evangelist''s
    proposal, not an i6eal product. It is deliberately collapsed: a naive 1-tool-per-operation mapping
    would emit 83 tools that differ only by filename, which is unusable for an agent.'
server:
  name: i6eal-open-ai-data
  transport: http
  url: null
  status: not published
design_note: 'Because there are no parameters anywhere in the spec, a faithful MCP surface has to introduce
  its own: the natural tool shape is (dataset, distribution) selectors resolved against the DCAT catalogue,
  which is exactly what the catalogue exists for. Each candidate tool below therefore names the operationIds
  it would dispatch to rather than pretending a parameterised endpoint exists.'
tools:
- name: list_datasets
  description: List all 16 i6eal datasets with identifier, bilingual title, description, temporal coverage,
    observation start, collection state and cadence, sources and rights.
  source_operation: openapi/i6eal-open-ai-data-api-openapi.json#getOpenDataCatalogue
  backing_url: https://i6eal.de/data/catalog/dcat.jsonld
  input_schema:
    type: object
    properties: {}
- name: get_dataset_index
  description: Fetch the full JSON index for one dataset — counts, generatedAt, trackingSince, schemaVersion,
    collectionStatus, sourceHealth, fingerprints and the named sibling distributions.
  source_operations:
  - getProcurementChronicleIndex
  - getProcurementObservedChangesIndex
  - getFederalAiBudgetMonitorIndex
  - getPublicAiSystemEvidencePassIndex
  - getGermanAiEvidenceNetworkIndex
  - getFederalAiCaseLawObservatoryIndex
  - getAiSupervisionMonitorIndex
  - getAiStandardisationMonitorIndex
  - getEuAiTransparencyEvidenceMonitorIndex
  - getGermanEuAiAdoptionAtlasIndex
  - getParliamentaryAiAccountabilityMonitorIndex
  - getGermanStateAiEvidenceLayersIndex
  - getEuPlatformAutomationMonitorIndex
  - getEuAiMedicalDeviceLifecycleMonitorIndex
  - getGermanyEuAiEnforcementMonitorIndex
  - getEuAiClinicalTrialLifecycleMonitorIndex
  input_schema:
    type: object
    required:
    - dataset
    properties:
      dataset:
        type: string
        description: dct:identifier from the DCAT catalogue, or the /data/<slug>/ path segment
- name: get_dataset_table
  description: Fetch one CSV table belonging to a dataset (RFC 4180, header row, UTF-8). 37 tables exist
    across the 16 datasets.
  source_operations_count: 37
  input_schema:
    type: object
    required:
    - dataset
    - table
    properties:
      dataset:
        type: string
      table:
        type: string
        description: Table name without extension, e.g. events, changes, titles, series, revisions, systems,
          links, entities, edges, decisions, passages, authorities, instruments, relationships, work-items,
          providers, evidence, countries, observations, questions, answers, platforms, flows, timeline,
          devices, actions, studies.
- name: get_dataset_methodology
  description: Fetch a dataset's methodology.json — how it is collected, which official hosts are allowed
    as sources, and what the dataset explicitly does not claim. An agent should read this before quoting
    any figure.
  source_operations_count: 16
  input_schema:
    type: object
    required:
    - dataset
    properties:
      dataset:
        type: string
- name: get_dataset_changes
  description: Fetch a dataset's Atom 1.0 change feed of observed material changes. Available for 11 of
    the 16 datasets (5 at atom.xml, 6 at feed.atom).
  source_operations_count: 11
  input_schema:
    type: object
    required:
    - dataset
    properties:
      dataset:
        type: string
coverage:
  rest_operations_total: 83
  candidate_tools: 5
  operations_reachable_via_candidate_tools: 81
  operations_not_covered:
  - getProcurementObservedChangesArchive
  - getGermanStateAiEvidenceLayersSchema
  operations_not_covered_note: Two one-off distributions (a rolling archive.json and one dataset's explicit
    schema.json) have no sibling on other datasets, so they do not fold into a generic tool; a get_dataset_distribution
    escape hatch keyed on the DCAT dcat:accessURL would cover them.
rating_note: 'No `type: MCPServer` pointer is wired into apis.yml for this file. The artifact is a candidate
  design derived from the OpenAPI, not a server i6eal operates, and an MCPServer pointer would award mcp_server_present
  and the agent-readiness mcp_server dimension for a surface that does not exist. Wire the pointer only
  if and when i6eal publishes a real endpoint that answers tools/list.'
agent_surface_today:
  llms_txt: https://i6eal.de/llms.txt
  openapi: https://i6eal.de/openapi.json
  dcat: https://i6eal.de/data/catalog/dcat.jsonld
  note: 'Until an MCP server exists, the working agent path is: read llms.txt for the operating rules,
    resolve URLs from the DCAT catalogue, then GET the distribution with a conditional request. No credentials
    are required at any step.'
deployment:
  mode: none
  verified: derived
  tools: 5
  checked: '2026-08-12'
  source: catalog MCP census