Docs for v3.1.0
One topic at a time. Pick a task, copy the command, then use the arrows or swipe for the next topic.
Install the CLI, index your repo once, then ask for context scoped to a task.
Builds the graph, memory and agent entry point in .tokencap/ so agents stop re-reading the repo.
TokenCap snapshot written: ./.tokencap/snapshot.md Knowledge graph written: .tokencap/graph/summary.md Memory layer written: .tokencap/memory/current.md
Install
Needs Node.js 18 or newer and Git.
Index the repo
TokenCap snapshot written: ./.tokencap/snapshot.md Knowledge graph written: .tokencap/graph/summary.md Memory layer written: .tokencap/memory/current.md Security engine: active (redaction engine v1)
Run from the project root. Writes everything to .tokencap/.
Ask for task context
Builds a ranked context pack for that task.
After tokencap make, agent pointer files such as AGENTS.md and CLAUDE.md send your AI tool to .tokencap/agent/START_HERE.md. Re-run tokencap make after big changes, or use tokencap make --watch to rebuild on save.
Upgrading from v2? Run tokencap make --migrate once, then tokencap make.