All features at a glance
This is the complete map. Every feature is listed once, grouped by what it does, with its default state and a link to the page that covers it. Use it to find a capability fast or to see the whole surface at once.
Default states use three values. On means it runs automatically after install with no action from you. Opt-in means it is off until you turn it on. On demand means it is always available but only runs when you invoke it.
Audit and visibility
Section titled “Audit and visibility”The commands that show you where context goes and what each session cost.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Full audit workflow | The /token-optimizer scan-fix-verify workflow that recovers structural waste | On demand | First audit |
| Full report | Per-component token breakdown with optimization suggestions (report) | On demand | Setup audit |
| Quick scan | Ten-second overhead, degradation risk, and top offenders (quick) | On demand | Setup audit |
| Dashboard | Single-file HTML dashboard: overview, quality, trends, manage | On (regenerates each session) | The dashboard |
| Trends report | Per-session usage, quality, cache, and cost over N days (trends) | On demand | Usage analytics |
| Savings report | Measured, estimated, and opportunity savings (savings) | On demand | Usage analytics |
| Turn breakdown | Per-API-call tokens, cost, model, and tools (conversation) | On demand | Usage analytics |
| Pricing tier | The pricing tier used for all cost math (pricing-tier) | Anthropic | Configuration |
| Dashboard daemon | Bookmarkable dashboard URL at a fixed port (setup-daemon) | Opt-in | The dashboard |
| Daemon and schema diagnostics | Probe the daemon and validate dashboard data (daemon-status, dashboard-diagnose) | On demand | Health and diagnostics |
| Security report | Enterprise security self-assessment (security-report) | On demand | Health and diagnostics |
Active compression
Section titled “Active compression”Runtime savers that cut new content as it enters your window. All on by default.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Bash output compression | Rewrites safe read-only CLI output to compact summaries, credential-safe | On | Bash output compression |
| Delta mode | Re-reads of changed files return only the diff | On | Read cache |
| Structure map | Large code re-reads return a function-and-class skeleton | On (soft-block) | Read cache |
| Read cache management | Clear or inspect the per-session read cache (read-cache-clear, read-cache-stats) | On demand | Read cache |
| Tool result archive | Large tool results saved to disk, retrievable with expand after compaction | On | Tool result archive |
| Active compression manager | Status and toggles for the compression suite (v5 command) | On demand | Active compression |
| JSONL trim and dedup | Remove oversized results and duplicate reminders from transcripts (jsonl-trim, jsonl-dedup) | Dry-run | Full CLI reference |
Smart compaction and continuity
Section titled “Smart compaction and continuity”Keep your work alive across auto-compact, crashes, and fresh sessions.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Checkpoint capture and restore | Saves session state before compaction, restores it after (compact-capture, compact-restore) | On | Smart compaction |
| Progressive checkpoints | Captures at multiple fill and quality thresholds, not just emergencies | On | Smart compaction |
| Milestone checkpoints | Snapshots before agent fan-out (checkpoint-trigger) | On | Smart compaction |
| Compact instructions | Mode-aware PRESERVE/DROP guidance injected at compaction time | On (dynamic) | Smart compaction |
| Cold-resume-lean | Reopen a stale session from its checkpoint for free (resume-lean) | On demand | Session continuity |
| Fresh-session nudge | Offers a clean restart when a long session degrades | On | Session continuity |
| Prompt continuity | Injects a topic-matched hint from a prior session on a new prompt | On | Session continuity |
| Checkpoint browser and stats | List checkpoints and view telemetry (list-checkpoints, checkpoint-stats) | On demand | Session continuity |
Quality scoring
Section titled “Quality scoring”A live grade for how healthy a session is, plus real-time alerts.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Quality analyzer | The 0 to 100, 7-signal context-quality score (quality) | On demand | Quality scoring |
| Status line | Context fill and quality grade in the terminal status bar | On | The quality status line |
| Quality nudges | Warns when quality drops suddenly so you compact at the right moment | On | Quality nudges and loop detection |
| Loop detection | Flags when the model is stuck retrying the same approach | On | Quality nudges and loop detection |
| Compression stats | Per-feature events, tokens saved, and quality preserved (compression-stats) | On demand | Quality nudges and loop detection |
| Validate impact | Statistical before-and-after check on session efficiency (validate-impact) | On demand | Health and diagnostics |
| Attention optimizer | Scores and reorders CLAUDE.md for the LLM attention curve (attention-score, attention-optimize) | On demand | Attention optimizer |
Coaching and fleet
Section titled “Coaching and fleet”Behavioral analysis across your history and across your agent systems.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Token Coach | Reads 30 days of history for setup and behavioral patterns (/token-coach) | On demand | Token Coach |
| Fleet Auditor | Scans every installed agent system for waste and dollar savings (/fleet-auditor) | On demand | Fleet Auditor |
| Waste detectors | The per-platform detectors for idle burns, misrouting, and config bloat | On demand | Waste detectors |
| CLAUDE.md injection | Inject data-driven coaching and model-routing advice as a managed block (inject-coach, inject-routing) | Opt-in | CLAUDE.md injection |
Cache and cost
Section titled “Cache and cost”Recover the prefix re-write penalty paid when a session resumes past its cache window.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Keep-Warm | Pings the cache before expiry so a resume stays warm (keepwarm-enable) | Opt-in | Keep-Warm |
| Keep-Warm forecast and backfill | Project savings from your own history before enabling (keepwarm-forecast, keepwarm-backfill) | On demand | Keep-Warm |
| Keep-Warm report | Mode, spend, realized savings, and tripwire state (keepwarm-report) | On demand | Keep-Warm |
| Subscription quota value | Measures Keep-Warm value in rate-limit quota for flat-rate plans (keepwarm-quota-value) | On demand | Keep-Warm |
| Cache TTL watchdog | Reports tokens re-written because a session resumed after cache expiry (cache-report) | On demand | Cache TTL watchdog |
Memory and configuration health
Section titled “Memory and configuration health”Keep CLAUDE.md, MEMORY.md, skills, and MCP servers lean and correct.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Memory review | Audits memory files for stale entries, size, and overlap (memory-review) | Read-only | Memory health |
| Skill and MCP toggles | Archive or restore skills and disable or enable MCP servers (skill, mcp) | On demand | Managing skills and MCP |
| Doctor | Verifies every component is installed correctly (doctor) | On demand | Health and diagnostics |
| Drift check | Compares the current setup against the last snapshot (drift) | On demand | Health and diagnostics |
| Snapshot and compare | Before-and-after snapshots of your setup (snapshot, compare) | On demand | Health and diagnostics |
| Git context | Suggests relevant files from current git state (git-context) | On demand | Health and diagnostics |
| Self-heal on start | Repairs drifted config on every session start (ensure-health) | On | Reference: hooks |
| Kill stale sessions | Ends sessions stuck in a running state (kill-stale) | On demand | Health and diagnostics |
Data and privacy
Section titled “Data and privacy”Control the local data Token Optimizer keeps.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Purge all data | Lists, then deletes all local data on confirm (purge) | Dry-run | Your data and privacy |
| Consent management | Shows, grants, or resets the data-notice consent flag (consent) | On demand | Your data and privacy |
| Daemon consent | Sets consent for the persistent dashboard URL (daemon-consent) | On demand | Your data and privacy |
Setup and platforms
Section titled “Setup and platforms”Install, hooks, and the per-platform adapters.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Plugin and hook install | Install the plugin and wire the hooks (setup-hook, setup-all-hooks) | On demand | Choosing your install |
| Smart compaction and quality bar setup | Install the compaction and status-line hooks (setup-smart-compact, setup-quality-bar) | On (plugin) | Choosing your install |
| Codex adapter | Install and run on OpenAI Codex (codex-install, codex-doctor) | Opt-in | Codex capabilities |
| Copilot adapter | Install and run on GitHub Copilot, beta (copilot-install, copilot-doctor) | Opt-in | Copilot capabilities |
| Hermes adapter | Install and run on Hermes, beta (hermes-install, hermes-doctor) | Opt-in | Hermes capabilities |
| OpenCode and OpenClaw | Native TypeScript plugins | Opt-in | OpenCode, OpenClaw |
Diagnostics and developer tools
Section titled “Diagnostics and developer tools”The deeper utilities for validating and tuning the compression engine.
| Feature | What it does | Default | Learn more |
|---|---|---|---|
| Benchmark | Runs the compression fixture suite (benchmark) | On demand | Benchmarks |
| Structure map inspector and proof | Preview or replay structure-map savings (structure-map, structure-proof) | On demand | Benchmarks |
| Cohort tripwire | Live edit-rate per cohort and auto-demotion state (cohorts) | On demand | Benchmarks |
| Context size override | Override the detected context window (--context-size) | On demand | Configuration |
For the complete command surface with every flag, see the Full CLI reference. For what runs without your intervention, see Automatic hooks.