knowledge-mapper — Claude Code subagent: what it does, how to install
knowledge-mapper turns a subject into a map of concepts: nodes, typed links, clusters and routes between ideas. Handy when a topic is too big to keep in your head at once.
no license ★ 1,357 updated 2026-03-01
.claude/agents/knowledge-mapper.mdWhat knowledge-mapper does
Every concept becomes a node carrying a domain, an importance weight and coordinates. Links between nodes get a type and a strength; the types run from is-a and part-of to causes, requires, opposite-of and leads-to. Groups of related nodes become clusters with a centroid and a coherence score.
You drive it with starred commands: *build-graph reads the current context, *map-connections lists what a concept touches, *trace-path connects two ideas, *find-gaps reports empty areas, *export-graph writes the result out. Activation runs /semantic:tree-init, loads stored maps from .wfgy/maps/, then prints *help. The agent runs on sonnet and can read, write and search files, run Bash, and fetch pages from the web.
Helper commands live in .claude/commands/semantic/ and .claude/commands/reasoning/. Skip them and activation's init step has nothing to call, because /semantic:tree-init resolves to a file in that first folder. The agent itself is one file: .claude/agents/knowledge-mapper.md.
What it helps with
- split a sprawling topic into nodes and links
- trace a route between two concepts
- lay out a study order for an unfamiliar field
- spot which parts of a knowledge map stay empty
- export a concept graph for a write-up
Example prompts for the agent
- Map the concepts in this spec: run
*build-graph, then*visualize-map. - How do
transactionsandisolation levelsconnect? Use*trace-path. - Run
*find-gapsover my notes and tell me what's missing.
How to install knowledge-mapper
Manually: put the files in place
.claude/agents/knowledge-mapper.mdFor all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention
Manually: put the files in place
.codex/agents/knowledge-mapper.tomlCodex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/
Manually: put the files in place
.cursor/agents/knowledge-mapper.mdCursor also reads .claude/agents/
Step by step, with commands for each platform: Claude Code subagents: how to create one, call it and keep context clean
Questions about knowledge-mapper
How to install knowledge-mapper in Claude Code?
Drop the agent file into your project at .claude/agents/knowledge-mapper.md. That path is the install for Claude.
What does knowledge-mapper do?
It builds a graph of concepts with typed, weighted links and clusters, then answers questions about routes, gaps and study order.
What does it need to work?
The command files under .claude/commands/semantic/ and .claude/commands/reasoning/, since activation calls /semantic:tree-init. Saved maps are read from .wfgy/maps/.