knowledge-mcp

Interactive exploration and requirements discovery using the knowledge store. Searches existing decisions, findings, and code before exploring new ideas. Records discoveries as research, findings, and decisions. Use when exploring options, discussing architecture, or investigating before planning.

Orchestration discipline. Defines the team hierarchy, your role as Engineering Manager, signal routing, drift detection, and the failure modes that make you a bad manager. Loads at the brainstorm-to-execute boundary and persists through ticket execution.

Introspect on your thought graph — examine personality, influence, tensions, blind spots, and reasoning patterns. Use for metacognition sessions, debugging reasoning, or understanding how your thinking has evolved.

Record an architectural or design decision in the knowledge graph with full rationale. Use after making a significant choice that future developers should know about.

Join a hive as a worker (claim work, do it, report the result) or act as a coordinator (dispatch role-targeted work and read the outcomes). A hive is a cloud work-queue for coordinating multiple agents across machines. Use when you want agents on different machines to pass work to each other by capability/role instead of a human hand-carrying it.

Research a topic using the knowledge store. Searches code, knowledge nodes, and existing decisions to build understanding. Use when investigating how something works, exploring options, or gathering context before implementation.

Capture the session feedback loop after work is verified for real — record a reproduction/interaction guide, charge the session's reasoning with the real-world evidence, record findings, and close the ticket. Use AFTER a feature is smoke-tested and confirmed working, or an investigation is remediated — never before.

Execute an implementation plan from the knowledge graph step by step. Updates status, verifies criteria, records thoughts about what you encounter, and charges thoughts when evidence arrives. Use after a plan has been created and approved.

Build live causal context about a repo. Authors non-trivial thoughts that answer WHY systems exist and behave the way they do, weaving evidence across code, cloud, practice, and knowledge graphs. Distinct from /research (which describes WHAT) and /improve (which surfaces what's wrong) — /explore answers why.

Ingest design patterns from an authoritative source (book, public catalog, reference site) into the practice/design-patterns.bin library graph. Two-phase workflow — recipe-driven bulk pass first (cheap, ~minutes/source), then optional hand-crafted refinement on selected patterns via the pattern-ingester agent (high-fidelity synthesis of use_cases / examples / references). Default to recipe-only and refine selectively; full agent runs are reserved for sources where every pattern needs decision-grade synthesis.

Create an implementation plan in the knowledge store. Researches the codebase first, then creates a structured phased plan with success criteria. Use when starting a new feature, refactor, or multi-step task.

Design a test plan collaboratively. Researches what needs testing, discusses scope and criteria interactively, then creates a structured test plan with steps and pass/fail criteria.

Execute a test plan from the knowledge graph. Starts a run session, spawns tester agents to run each test step, and reviews pass/fail/skip results.