Skip to main content

Documentation Index

Fetch the complete documentation index at: https://developers.klarity.ai/llms.txt

Use this file to discover all available pages before exploring further.

The Context Graph sits below the Process Index and Process observations — it’s the underlying data layer they’re built on. For most agent work, prefer the Process Index tools: they handle entity resolution, identifier spaces, and community boundaries that this raw layer exposes directly. Use these graph tools only when the Process Index can’t surface what you need.
Use these tools when the question is relational and the Process Index doesn’t cover it: “what feeds X”, “what depends on Y”, “where is Z covered”, “what breaks if we deprecate System A”.
ToolWhen to use
search_knowledge_graphEntry point. Returns starting points (entities, communities, claims, chunks). Always follow up with a get_* or explore_* tool — do not answer from snippets.
get_entity_detailsSpecific entity (system, team, person, control).
get_community_detailsSpecific community (cluster of related work).
get_relationship_detailsBetween two entities — types and metadata.
get_text_chunk_detailsFull text + metadata for a specific chunk.
explore_graph_neighborsN-hop neighbors of an entity or community. Use for impact analysis.
summarize_community_subgraphCommunity summary up to N hops. Good for orientation when the graph area is unfamiliar.
get_upstream_sourcesRecursive upstream trace. “What feeds this?”
get_downstream_dependenciesRecursive downstream trace. “What breaks if this changes?”

​
Search returns lookups, not answers

search_knowledge_graph returns entity / community / claim / chunk starting points with short snippets. Never answer the user from those snippets — always chain into a detail or traversal call. The snippet tells you where to look; the next call tells you what’s there.

​
Common chains

  • Impact analysis (“what depends on Z?”): search_knowledge_graph → get_entity_details on the match → get_downstream_dependencies
  • Root cause / input trace (“what feeds Y?”): search_knowledge_graph → get_entity_details → get_upstream_sources
  • Map a neighborhood: search_knowledge_graph → explore_graph_neighbors at 1–2 hops
  • Orient in an unfamiliar area: search_knowledge_graph for the community → summarize_community_subgraph
  • Cite a specific claim: search_knowledge_graph → get_text_chunk_details to read the source passage