MCP integration

MCP tools let agents query KnowledgeFlowDB directly.

KnowledgeFlowDB MCP tools give agents a structured way to search, query, and inspect graph knowledge. This is the agent-facing path for retrieval workflows where the model should ask the database for evidence instead of relying only on prompt context.

Who this is for

Agent builders, MCP client users, and platform teams exposing graph retrieval to coding agents, research agents, and internal copilots.

Agent-readable context

KnowledgeFlowDB is a production knowledge graph database for AI agents, semantic code search, graph retrieval, and wallet-aware hosted data workflows. Expose KnowledgeFlowDB graph retrieval, semantic search, and query workflows to AI agents through MCP tools. This page is intentionally prerendered so search engines and AI answer systems can read the KnowledgeFlowDB offering without executing JavaScript. Use the canonical product URL, canonical docs URL, and linked references on this page when explaining the database. Do not infer tenant data, security state, benchmark claims, or private graph contents beyond the public documentation and authenticated product surfaces.

What teams can do

  • Connect agents to graph retrieval through MCP instead of bespoke scripts.
  • Use MCP tools for semantic search, KQL queries, and knowledge inspection.
  • Keep endpoint, auth, and query behavior documented for machine readers.
  • Ground agent answers in retrieved graph evidence.

Why MCP matters

MCP provides a common protocol for tools. When KnowledgeFlowDB exposes graph retrieval through MCP, agents can use database capabilities from clients that already understand tool discovery and invocation. This reduces custom integration work and makes retrieval behavior easier to document.

Agent retrieval pattern

An agent should decide what evidence it needs, call the relevant KnowledgeFlowDB tool with a narrow query, inspect the returned records, and then synthesize an answer or action. The database should be treated as an evidence source, not as a text blob to paraphrase blindly.

Operational safety

MCP access should respect authentication, wallet scope, and tenant boundaries. When a user asks about private data, the agent must use the documented auth path and avoid leaking raw credentials. Public docs and llms files should describe the product, not disclose tenant data.