01
See what exists
Understand detectable tools and supported usage surfaces, with clear distinctions between observed and unknown.
DISCOVERY IN ACTIVE DEVELOPMENT / LOCAL FIRST
Make sense of
your AI usage.
When your AI tools are scattered, your usage story shouldn't be. UsageHarbor brings evidence about activity, capacity, and wasted effort into one privacy-minded Windows cockpit.
Early-stage product · No public installer or open pilot signup yet
CONTROL ROOM / 001ILLUSTRATIVE SAMPLE
YOUR WORKSPACEOverview
Local evidenceObserved sources04Sample only
Capacity signals02Source linked
Review items03Human decision
Activity at a glanceEXAMPLE DATA
Evidence has a sourceObservations, not invented balances
OBSERVE FIRST. ACT DELIBERATELY.WINDOWS / LOCAL
Built around what you can verifyLOCAL DATASOURCE PROVENANCEHONEST UNKNOWNSHUMAN CONTROL
01 / THE PROBLEM
More AI tools.
Less clarity.
Subscriptions, command-line tools, background workflows, rate limits, and usage logs each tell a different part of the story.
A percentage without a timestamp isn't reliable evidence. A cost estimate isn't a bill. And a full quota meter doesn't tell you why work was repeated.
UsageHarbor connects the pieces it can actually observe.
02 / THE APPROACH
A clearer picture.
Without the guesswork.
One local-first workflow, built around evidence and careful boundaries.
02
Follow the evidence
Keep observations attached to their source and original time. Stale, absent, or incompatible values remain explicit.
PROVENANCE03
Spot wasted effort
Review activity patterns, repeated work, and potential capacity pressure without manufacturing false savings.
ATTRIBUTION04
Stay in control
Get evidence-backed recommendations to review. UsageHarbor does not choose or dispatch your AI models.
DECISIONSFeatures described here represent the active product direction and source-level implementation, not a promise that every provider or workflow is already supported.
03 / OUR PRINCIPLES
Your usage data
is yours.
Local-first is more than a label. The goal is useful answers without uploading your private conversations just to understand what your tools are doing.
01
Observe, don't impersonate
Read supported evidence. Don't invent provider permissions, identities, prices, or quota guarantees.
02
Keep unknowns visible
Missing or expired source data stays unknown. An estimate never silently becomes a fact.
03
Leave execution to the operator
UsageHarbor describes usage and recommends changes. It doesn't autonomously launch or route AI workloads.
04 / THE BUILD
Being built.
Carefully.
UsageHarbor has a tested Windows source foundation, including local capacity evidence and analysis. Public distribution is a separate milestone.
Read the FAQSource foundationsLocal implementation and automated source checks
VERIFIEDProduct and pilot validationUsability, provider coverage, recovery and real-world feedback
IN PROGRESSTrusted public Windows releaseSigning, isolated installer tests, and distribution review
NOT RELEASED05 / THE DETAILS
Good questions.
Can I download UsageHarbor?
Not publicly yet. Source-level checks and internal development builds are not the same thing as a trusted, tested public installer.
Which systems are supported?
Windows is the current development target. A public release and Mac or Linux support are not currently promised.
Does it switch models automatically?
No. UsageHarbor reports evidence and recommendations. Provider admission and dispatch belong to a separate operator system.
Does it read my conversations?
Its design prioritizes metadata and bounded observations rather than retaining prompt or conversation bodies. Individual data-source support is still being validated.