Local AI workbench
Runs against the published graph in this browser. Results are visibly rendered as JSON and preserve the authored edge direction, relation metadata, citations, and content version.
Inputs accept canonical IDs. Search first when a name is uncertain. Relation-type restriction is optional; leave it unselected to use all authored relation types.
Search concepts
Resolve names, IDs, taxonomy, and authored text to canonical mAtlas concepts.
Direct relations
Return direct incoming and outgoing typed authored relations for one concept.
Find authored paths
Find shortest paths through existing authored edges. Reverse traversal never reverses the original edge assertion.
Compute closure
Compute the distinct predecessor or prerequisite closure defined by relation-type metadata.
Connect concepts
Connect several concepts with a deterministic shortest-path heuristic over authored edges.
Build a bounded subgraph
Expand direct neighbors around one or more concept IDs for a compact, machine-readable context package.
Compare concepts
Compare authored metadata, direct relationships, common fields, domains, sources, and neighbors.
Create canonical links
Return a canonical concept URL and an interactive-graph URL for one concept.
Machine-readable result
A browser agent can read this result directly from the DOM. Use Copy JSON to move it into another AI context.
{}