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.

ToolWhen to use
get_schemaRead the workspace database schema.
research_webResearch a topic on the public web. Use sparingly — Klarity’s value is grounded in the customer’s workspace, not the open web.
A raw SQL execute_query tool exists for advanced fallback in some configurations, but is not exposed by default. Prefer typed tools over SQL unless the question genuinely cannot be answered any other way.

​
Schema as last resort

get_schema returns the workspace PostgreSQL schema — table and column structure. Use it when:
  • The user asks an analytics question no typed tool covers.
  • You’re generating code that needs the data model (Build agents, skills, and code).
  • You want to confirm field semantics before building a downstream deliverable.
Don’t reach for the schema for everyday process questions. The typed tools above are faster, safer, and produce more interpretable results.

​
Web research

research_web pulls external context — published frameworks, vendor positioning, regulatory guidance. Use it as the second source in a gap analysis (the workspace is the first source), not the primary answer. See Combine with deep research for the pattern.

​
Anti-patterns

  • Don’t answer a process question with research_web before checking the workspace. The whole value of the MCP is that the workspace is the ground truth.
  • Don’t call get_schema to answer “what does our vendor onboarding look like” — use search → fetch.
  • Don’t treat raw SQL output as a citation. Cite the typed object (process ID, observation timestamp, artifact ID) the user can audit in the Klarity UI.