Capybara Broker
Makes every token count
A calm approach to reusable context and efficient token flow.
From2026-04-13to2026-10-13
7D30D90D6M1Y
DeviceAll devices
AgentAll agents
ModelAll models
Total tokens121.59B121,593,247,261 lifetime
Cost USD$92340.77$92340.77 total
Total input121.16B121,164,192,903 input
Output532.97M532,972,070 output + reasoning
Cache hit rate95.40%115,586,819,747 cached / 121,164,192,903 total input
Reasoning output115.59M115,587,017
Last 7 days5.25B / $3310.645,247,681,894 tokens / $3310.64
Last 30 days12.22B / $9700.6312,220,066,884 tokens / $9700.63
Daily Trend
Recent 30 days · 12.22B
Track Token Flow
Record AI usage at the source, then turn token flow into cost, rhythm, and role insight.
Role Insights
Your usage is matched to one of 10 roles
TokenFlow looks at your last 30 days, such as when you work, how many agents and model vendors you use, and how much context you reuse, then picks the role that fits best. These are all of them.
LV.3
Swallow Courier
Carries work between tools
Moves between harnesses and keeps the work moving.
Penguin Engineer
Builds a dependable workbench
Steady sessions and connected devices form a dependable workshop.
Rooster Timekeeper
Sets the morning rhythm
Gets work moving when the morning arrives.
Owl Sentinel
Keeps the night watch
Finds a quiet rhythm in the late-night hours.
Tabby Bard
Turns tokens into expression
A larger output share gives ideas room to become words.
Fox Alchemist
Mixes models and possibilities
Explores different model providers and combines their strengths.
Wolf Weaver
Weaves tools into one flow
Several harnesses work together like threads on a loom.
Hedgehog Painter
Gives ideas a visual form
Image and video model activity brings a visual rhythm to the work.
Rabbit Scribe
Keeps context close at hand
Returns to familiar context and develops it with care.
Highlights
Live sharing and embeds
Shared views follow your dashboard
Every README card and blog chart below comes from this dashboard. When new usage syncs, each shared copy updates with it, so there is nothing to re-export.
README usage blocks
Paste one Markdown line into a GitHub profile; the refreshable SVG shows activity, streaks and active days.
Blog-ready activity views
Embed trend, hourly rhythm, cost or model-share charts in a post; readers hover for daily detail.
Capture Mesh
One account, every path into AI.
Devices, harnesses, and providers rotate throughout the day. TokenFlow keeps every stream attached to the same identity.
Usage metadata only. No prompt text.
TokenFlow syncs time-bucketed tokens, estimated cost, model, agent, provider, and device labels. Prompts, responses, files, API keys, and local content stay off the upload path.