Frequently Asked
Questions

A source-reviewed catalogue of MCP servers, skills, libraries, and agents that may support real-world evidence workflows. It adds an RWE interpretation layer to public project information: workflow fit, documented capabilities, limitations, operational scope, and evidence state.

The website displays the Core catalogue. The expansion gathered capability records across MCP servers, skills, libraries and components; the total was never a count of unique MCP servers. A subsequent relevance audit separated adjacent discoveries and unresolved scope questions into dedicated Excel sheets. Those records are kept out of the website catalogue.

Black Swan Causal Labs maintains the registry as an independent, living research resource. Editorial judgments are based on publicly available sources and are revised when the underlying evidence changes.

A public MCP server, connector, agent skill, library with an agent-facing interface, plugin, or orchestrated agent workflow must support an identifiable RWE, HEOR, epidemiology, public-health, evidence-synthesis, or specific statistical or evidence-synthesis function. A merely possible health application, generic productivity function or molecular-only workflow does not qualify for the core. It must also have enough public evidence to describe its identity and function responsibly.

The registry can observe only publicly discoverable work. Private repositories, internal enterprise systems, unpublished tools, access-controlled packages, and public projects with insufficient searchable metadata may not be found. Absence means “not found in the searched public sources,” not “does not exist.”

No. Inclusion is not a security, privacy, clinical, methodological, regulatory, or quality certification. A project may be relevant to an RWE workflow and still require substantial technical, governance, and subject-matter review before use.

No. The relevance pass individually reviewed the existing capability and limitation descriptions, with selected source follow-ups identified on the records. It did not repeat every primary-source review, install projects, or rerun the discovery campaign. Source-check, capability-review and relevance-review dates have different meanings.

Public repository files, documentation, manifests, packages, or listings were examined without running the project. Static review can confirm documented implementation and scope, but it cannot establish that installation succeeds, tools behave as claimed, outputs are correct, or sensitive data are protected.

No candidate projects were installed, invoked or runtime-tested in this continuation. Cards record identity, metadata, documentation or selected implementation review. Any earlier runtime evidence must be stated explicitly in its record; inclusion does not establish that a tool works correctly.

Added is the original catalogue inclusion date. Checked records the latest stated source check: historical cards may show an availability check, while continuation cards record completion of editorial source review and reconciliation. Expanded cards separately show the capability-review date. An availability check does not renew that review or validate tool behavior. A gold star identifies a Black Swan Causal Labs product, not an independent quality or validation rating.

Workflow-fit labels are concise editorial translations of documented capabilities. In the archive they preserve earlier discovery context and may not justify core inclusion. In the core, the relevance rationale identifies the retained function; they do not establish fitness for a particular study, dataset, decision, or regulatory use.

Collection-level claims remain at collection level. A capability belonging to one constituent package or skill is attributed to that component, not to the entire collection. Components may receive separate records when the additional detail materially improves discovery.

Public-source discovery and surveillance are designed to run weekly. New or materially changed projects enter a private review queue. Approved records are added to the website and public workbook in reviewed batches rather than being published automatically.

The protocol combines structured registries, GitHub repository and code searches, curated catalogues, predefined RWE terminology and links from discovered projects. The September continuation expanded the search under a revised source-specific plan. Ranked directory searches, unavailable sources and the original uncompleted broad query matrix remain coverage limits. The Methods page and coverage report distinguish completed searches, automated screening and individual source reviews.

Canonical repositories, packages, aliases, redirects, and cross-listings are compared before a record is treated as new. Missing projects remain in the history as unavailable rather than being silently deleted, and reappearing projects are flagged for review.

Yes. Submit the canonical public source and a short explanation of its RWE relevance through the Black Swan Causal Labs contact form. Suggestions enter the same review process as discovered candidates and are not guaranteed inclusion.

Yes. The public dataset and original editorial content are licensed under Creative Commons Attribution 4.0. Website source code is licensed under MIT. Third-party names, trademarks, source descriptions, and linked materials remain subject to their respective owners’ rights.