Static demoMock data - nothing here is live. Install enigma and run enigma dash for the real, interactive control panel.Back to siteInstall Enigma
Enigma.
Control panel
Connecting...
Money Saved
-
Estimated, by source
Time Saved
-
Estimated prefill time
Tokens Saved
-
Across all compressions
Savings Rate
-
Of input tokens removed
Compressions
-
Reversible via CCR
Avg / Compression
-
Tokens saved per call
Best Compression
-
Most saved in one call
Enigma Systems what's active - configuration, not savings
Only Context Compression and (opt-in) real tool-usage are measured for savings below. The rest is shown as state: enigma can't measure their effect because they run inside the agent, not through enigma - so it reports their configuration honestly instead of inventing a number.
Loading...
Savings Per Day
No compression activity recorded yet. Run enigma compress <file> or enable the compression MCP (enigma config compress on).
Savings By Source which app/CLI compressed
Savings By Content Type what enigma compressed
Before vs. After cumulative tokens
Before
-
After
-
Removed -
Reduction -
Compression ratio -
Savings History
Recent Compressions newest first
Reversible Cache (CCR) recoverable originals on disk
Claude Usage
Measured token consumption and estimated cost, read-only from your own Claude Code session transcripts (~/.claude/projects). Cost is an estimate from a per-model price table; real spend is billed by Anthropic. Nothing is sent anywhere.
Cards marked live show the real % and reset Anthropic reports (the same as Claude's own UI). Turn on enigma config usage-api on to fetch them with your local Claude Code login - no proxy or plan limit needed. (The proxy also captures them from enigma claude traffic.) Until live data lands, the card falls back to tokens used, or a % against a plan limit you set (on the card, or enigma config plan-weekly-limit <tokens>). Per-model live data appears only when Anthropic sends it.
Est. Cost
-
All sessions, per-model pricing
Cache Saved
-
Est. from prompt caching
Input Tokens
-
Consumed across sessions
Output Tokens
-
Generated by the agent
Cache Reads
-
Tokens served from cache
Sessions
-
Transcripts scanned
Current 5-hour block
Window
-
Tokens -
Cost -
Burn rate -
Projected -
By Account per Claude login
By Model
By Project
Recent Sessions
Real tool usage opt-in
Turn on Real tool-usage stats in Settings to read your Claude Code sessions for token consumption and cost. enigma reads only your local session transcripts; nothing is sent anywhere.
No local usage data
Accounts & Profiles
Manage per-tool logins and profiles. A profile pins one account per tool and drives every launch. Logging an account in happens in a terminal - the browser shows you the command to run.
Accounts
Loading accounts...
Profiles
Loading profiles...
Skills
Engineering policy skills deployed to your agents - enigma's own, kept current automatically, plus any external skills found in your agent directories.
Editing
Skill settings
Agent memory
Customize the instruction files enigma deploys to each agent (CLAUDE.md / AGENTS.md). Your edits follow the Customized-skill updates policy on the next sync: overwrite restores the managed version, keep preserves them.
Loading...
Loading skills...
Settings
Everything you can configure with enigma config or the terminal UI, editable here. Changes apply at global scope and take effect immediately; toggles that change agent memory need an agent restart.
Loading settings...
Dashboard port
the port this dashboard binds. Leave 0 for auto (tries 80, then 24282, then an ephemeral port). Changing it needs a dashboard restart.
Preferred port
loopback only; never network-facing
Resources
Free up your machine: shut down WSL (the vmmemWSL RAM hog), quit Docker Desktop when it won't, free a stuck port, or kill any process. These actions are destructive and ask before running.
System
Free a port
Listening ports
Loading...
Processes by memory
Loading...
Recall
Searchable memory of your past coding sessions, built locally from your own transcripts. Agents can query it over MCP (enigma_recall); it never leaves this machine.
Loading...
Memory provider
Which model curates your memory - it keeps only what's worth remembering. The default reuses your local Claude Code login (no key). Or pick Anthropic or an OpenAI-compatible API and add its key. The key is stored encrypted, and ENIGMA_RECALL_API_KEY overrides it.
LLM curation
Off falls back to a local heuristic - no network calls
Search
Add a memory
Write one yourself, or generate it from a short note with the configured model.