graph-explore

Extract architectural context by querying the Graphify knowledge graph to summarize modules and dependencies into a concise, structured output. This approach limits tool calls and routes them through a subagent to deliver a ≤300-token summary that directly informs planning phases.

Yes, you can use a subagent to fetch and synthesize codebase dependencies for BMAD phases. The subagent queries the Graphify MCP server tools and returns a ≤300-token architectural summary, ensuring safe operation by preventing direct main-session MCP calls.

The best way to get a concise architectural summary is to query a knowledge graph for module and dependency context. By routing MCP tool calls through a subagent, you synthesize findings into a ≤300-token output, keeping the main session clean and focused for planning.

No, Graph Explore does not require direct MCP server calls in the main session. It enforces safe operation by routing all Graphify MCP tool interactions through the subagent, which then returns the final ≤300-token architectural summary to the main session.

Limiting architectural summaries to 300 tokens during BMAD phase planning ensures the output remains concise and immediately actionable. This constraint forces the subagent to synthesize only essential module, dependency, and structural context from the knowledge graph, preventing context overload.