generated: '2026-08-29' method: searched source: >- https://www.sparxsystems.jp/help/17.0/oslc_am_top.html, https://www.sparxsystems.jp/help/17.0/query_capability.html, https://www.sparxsystems.jp/help/17.0/creation_factory.html, https://www.sparxsystems.jp/help/17.0/oslc_serv_provid.html, https://www.sparxsystems.jp/help/17.0/info_accessed_via_oslcam.html, https://www.sparxsystems.jp/en/MCP/ description: >- Cross-cutting request/response semantics for the two programmatic surfaces an integrator can actually call: the Pro Cloud Server OSLC Architecture Management 2.0 RESTful API, and the STDIO MCP server. Enterprise Architect publishes no OpenAPI, so every convention below is read from the provider's own documentation rather than derived from a spec. surfaces: http: name: Pro Cloud Server OSLC Architecture Management 2.0 RESTful API deployment: customer-installed; base host is the customer's own Pro Cloud Server base_url_template: ':////oslc/' example_from_docs: 'http://localhost:480/firebird_model/oslc/' local: name: MCP Server for Enterprise Architect transport: stdio deployment: Windows add-in inside a licensed Enterprise Architect install endpoints: - path: '/{model_name}/oslc/am/login/' method: POST purpose: Obtain the user authentication token. - path: '/{model_name}/oslc/am/logout/?useridentifier={token}' method: GET purpose: Invalidate the token. - path: '/{model_name}/oslc/am/sp/' method: GET purpose: >- Service Provider Resource — the machine-readable discovery document. It names the URL to POST to in order to create resources, the URL to GET existing resources, and the URLs for the resource-shape metadata describing both the creation payload and the resource representation. This is the closest thing to a machine-readable contract the provider publishes, and it is served by the customer's own deployment. - path: '/{model_name}/oslc/cf/' method: POST purpose: Creation Factory — create a new resource by POSTing an RDF/XML representation. - path: '/{model_name}/oslc/qc/' method: GET purpose: Query Capability — query the model, returns RDF/XML. - path: '/{model_name}/oslc/re/{resource_GUID}/' method: GET purpose: Individual resource by GUID. This URL appears as rdf:about on each queried resource. authentication: style: token obtained by form POST, passed as the useridentifier parameter optional_shared_secret_header: EAO-Access-Code see: authentication/sparx-enterprise-architect-authentication.yml media_types: request: application/rdf+xml response: application/rdf+xml note: >- RDF/XML only. No JSON representation is documented. Namespaces in use are rdf, oslc_rm (http://open-services.net/ns/rm#), dcterms (Dublin Core Metadata Element Set 1.1), foaf, and the vendor namespace ss (http://www.sparxsystems.com.au/oslc_am#). query: style: OSLC query parameters on the Query Capability URL parameters: - oslc.where - oslc.select - oslc.properties - oslc.prefix pagination: documented: false note: >- No paging parameter, page-size limit, or continuation token is documented for the Query Capability. An integrator querying a large repository has no published way to bound a result set other than narrowing oslc.where. field_selection: supported: true mechanism: oslc.select and oslc.properties (OSLC Selective Properties conformance) metadata: tagged_values: >- Enterprise Architect tagged values are first-class and are readable, creatable, updatable and deletable through OSLC on packages, elements, attributes and operations. request_tracing: request_id_header: none documented versioning: scheme: product release train current: >- OSLC Architecture Management v2.0; served by Enterprise Architect 17.x with Pro Cloud Server 6.x. The API version is not carried in the URL path — /oslc/am/ is fixed and the behaviour tracks the installed Pro Cloud Server build. provider_statement: >- The provider states OSLC Architecture Management v2.0 functionality will continue to be improved, with changes announced in the release notes. error_envelope: format: RDF/XML error response (OSLC Error Responses conformance) see: errors/sparx-enterprise-architect-problem-types.yml rate_limit_signaling: documented: false see: rate-limits/sparx-enterprise-architect-rate-limits.yml idempotency: supported: false note: >- No idempotency key, no request-deduplication header, and no client-supplied request identifier is documented for the Creation Factory or for any OSLC update. A retried POST to /oslc/cf/ creates a second element. No Idempotency pointer is emitted. dry_run_mode: supported: false note: No preview/validate-only mode is documented on any write path. reversibility: grade: documented applies_to: write note: >- Enterprise Architect is a write-heavy modelling repository, so reversibility is the sharpest question an agent can ask before acting on it — and the answer differs starkly between the two surfaces. write_surfaces: - surface: MCP server (create/update/delete tools) reversal: operation: apply_baseline mechanism: >- Capture a package baseline with create_baseline before the agent acts, then restore the package with apply_baseline. This is a real, first-party, tool-level undo path exposed over MCP. window: null window_note: >- No retention window is stated. A baseline persists in the model until deleted, so the practical window is "as long as the baseline is kept" — but the provider does not state this, so it is recorded as unstated rather than unlimited. audit_trail: operation: '-modifiedInfoPath ' note: >- Launch flag that appends every AI add/edit/delete to a CSV, so AI-authored changes can be identified after the fact. An audit trail, not a reversal. provider_warning: >- The provider explicitly warns that deletion is not reliably scoped: during their own testing an instruction to delete a package's tagged values deleted the entire package. Both -enableEdit and -enableDelete are off by default, and the provider recommends backing up the project before enabling either. evidence: https://www.sparxsystems.jp/en/MCP/ - surface: Pro Cloud Server OSLC (Creation Factory, partial update, resource removal) reversal: operation: null mechanism: none published window: null note: >- OSLC supports Resource Removal and Partial Resource Update, but no cancel, undo, restore or soft-delete path is documented, and the provider states that OSLC updates are NOT written to the Enterprise Architect audit log even when auditing is enabled. An agent writing over OSLC therefore has neither a reversal operation nor an audit record of what it changed. This is the single most important runtime fact in this file. evidence: https://www.sparxsystems.jp/help/17.0/info_accessed_via_oslcam.html constraints: - >- Creation Factory requires the OSLC properties title and packageID in the RDF payload, and the packageID must name a package that already exists in the target model. A client must run a Query Capability request first to obtain a valid packageID. - >- The OSLC properties identifier, created and modified are read-only and cannot be set. - >- An empty value for a constrained property (the docs use ss:difficulty) is an error, not a no-op. - >- Element type strings passed to the MCP creation tools must exactly match the Automation Interface type names; the server ships MCP prompts (UML, SysML 1.5, BPMN 2.0) specifically to supply those strings to the model. cross_links: authentication: authentication/sparx-enterprise-architect-authentication.yml errors: errors/sparx-enterprise-architect-problem-types.yml lifecycle: lifecycle/sparx-enterprise-architect-lifecycle.yml rate_limits: rate-limits/sparx-enterprise-architect-rate-limits.yml data_model: data-model/sparx-enterprise-architect-data-model.yml mcp: mcp/sparx-enterprise-architect-mcp.yml