Tools & agents

Runs 2-4 attempts at a task in parallel git worktrees, each with a different approach; compare their diffs and adopt the best.

Finds the commit that broke something: git bisect with your test command in a temporary worktree (your working tree is never touched), live progress in a pane, the culprit with its diff stat handed to Claude.

Gives Claude a screen to draw on: SVG diagrams and Markdown reports in a Canvas pane with history.

More than one mind: ask several models at once and get a synthesis, critique the last answer, and let Claude consult another model.

Look into CSV, TSV, JSON and JSON Lines files without dumping them into context: schema, stats, rows, filters, group counts and single-row lookups, exact and compact, streamed with caps; a /peek pane with column sparklines.

An HTTP tool for API work: status, timing, trimmed headers and a body shaped to fit; localhost runs freely, other hosts ask once a session; an /http pane of the session's requests with replay and copy-as-curl.

A live plan Claude posts and keeps current: progress above the prompt, the full checklist and its history in a pane.

Search this project's past sessions: /recall for you (attach an old exchange to your next prompt), a tool for Claude to look up earlier decisions.

A compact, cached map of the repository for Claude: files by folder with sizes and each source file's symbols with line numbers, sized to a token budget, instead of reading files to get its bearings.

Runs your dev server or watcher for the session, flags its errors above the prompt, and lets Claude read its output.

CI that fixes itself with your say-so: watches the branch's workflow run after a push, puts a failure with its error-dense log on the band with Fix it, surfaces new review comments, and can pre-review PRs you are asked to review.

Works your backlog while you are away: when idle, one worker at a time implements the next .claude/backlog.md task in its own git worktree under a strict tool policy and budget; in the morning you review, adopt or discard each branch.

Runs your project's checks after Claude edits, shows pass/fail above the prompt, and closes the loop with one-press (or automatic, bounded) fixes.

Jev (TypeSafe's System One model) as Claude's instant co-processor: bulk filter/classify/rank/semantic-grep tools, risk notes under permission dialogs, a check on unverified 'it passes' claims, and optional per-prompt effort.

Opt-in: questions about the codebase (where is, what does, how does, which file, find) go to a read-only haiku scout subagent first; its answer comes with Ask Claude anyway and Good enough, and the tokens saved.