Context engineering toolkit for AI agents. 24 MCP tools across retrieval, memory, context, and team telemetry. .NET 10. No cloud deps.
Agent Claude Code
Use proactively when the user wants to plan a token budget, pack a system-prompt + retrieval + memory + team-context bundle into a budget, count tokens, or optimize for prompt caching. May read from retrieval and memory tools to assemble content. Does NOT index, store memories, or score teams.
Agent Claude Code
Use proactively when the user wants to index code or documentation and search it with Koshi's BM25 retrieval. Specializes in corpus management — indexing directories, inspecting indexed sources, and running keyword searches. Does NOT manage memories, compile context windows, or score teams.
Agent Claude Code
Use proactively when the user wants to store or retrieve facts, decisions, patterns, or preferences across Koshi sessions, capture decisions made in a turn, sync to/from a Markdown vault, or prune the memory store. Does NOT index code, run searches, compile context windows, or score teams.
Agent Claude Code
Use when the request spans multiple Koshi pillars (retrieval + memory, index-then-pack, score-then-tune), or the right specialist persona is not obvious. Generalist with access to all 24 Koshi MCP tools. Defers to single-pillar specialists (koshi-librarian, koshi-memory-keeper, koshi-context-packer…
Agent Claude Code
Use proactively when the user wants to score AI interactions, track per-team quality over time, render team dashboards, or ask "how should we tune our context engineering?". Does NOT index, store memories, search the corpus, or compile context windows.