Most habit trackers passively count checkboxes. Habitscore establishes your personal historical baseline, detects statistical deviations, recommends targeted micro-interventions when friction strikes, and learns from your response to adapt the next intervention.
Live API endpoint: api.mydev-tools.tech
A habit tracker that only tells you "you missed 3 habits" is useless when resistance hits. Habitscore implements a closed-loop system:
"You completed 2 of 5 habits today. Your streak is broken. Try harder tomorrow."
(Static, judgmental, ignores context, offers no actionable adaptation.)
"Your evening completion dropped 42% below your personal baseline. Late-day fatigue is correlated. Would you like to try a temporary 5-minute version today? Historically, task shrinkage recovered consistency by 82%."
(Personalized, context-aware, testable micro-intervention, closed-loop feedback.)
Traditional apps judge users against arbitrary, generalized population standards. Habitscore's Adaptive Behavioral Pattern Engine compares your current behavioral signature strictly against your own rolling statistical baseline.
Computes rolling mean, median, IQR (interquartile range), and standard deviation over 14–60 day windows to capture your natural rhythms and variances.
Calculates z-scores and quartile boundaries. Flags significant dips or spikes only when deviations cross statistical confidence thresholds (confidence ≥ 75%).
Extracts compact daily vectors: completion rate, postponement rate, duration, energy deltas, mood shifts, and temporal entropy.
Every user-facing report is structured around five empirical stages. This creates an objective feedback loop rather than arbitrary AI speculation:
Objective facts relative to personal baseline: "Your workout completion averaged 40% over the last 4 days, a 45% drop relative to your 30-day baseline (85%)."
Recurring context identification: "Habits scheduled or attempted after 19:00 show a 3x higher postponement rate."
Open, non-causal hypotheses: "Schedule friction, energy depletion, or task duration friction on high-workload days. Tracking sleep timing will reduce uncertainty."
One low-friction micro-action selected by contextual multi-armed bandit: "Shrink your routine from 30 minutes to 5 minutes today. Priority is continuity, not volume."
Outcome measurement: "You attempted and completed the 5-minute version with reported ease (difficulty: 2/5). Bandit weights updated: task-shrinkage success rate increased to 84%."
Test your own hypotheses with structured 7-day variant trials:
"Does reducing my target from 30 minutes to 10 minutes improve weekly consistency?"
The system measures baseline vs treatment periods and reports the measured delta without making global claims.
User-controlled summary export for therapists, doctors, or wellness coaches:
Generates clean adherence histograms, observed trend graphs, and intervention response tables.
Strictly evidence-only with zero clinical or psychiatric labels.
Not a Diagnostic Tool: Habitscore does not diagnose, treat, or cure any medical or mental health condition. It learns behavioral signatures relative to your personal historical baseline—not neural signatures or clinical classifications. We strictly separate Observation from Hypothesis, and Correlation from Causation.
Causal confidence is explicitly capped at 35% (pcause ≤ 0.35). The AI will never assert that "stress caused your lapse" or diagnose depression.
Raw journal notes stay local in the database. Only de-identified derived statistical vectors ever cross model boundaries.
Built with an extensible signal interface (SignalSource → SignalNormalizer → FeatureExtractor) ready for future wearable and sensor streams.
Published in a machine-readable manifest so an orchestrator or AI agent can match on capability:
behavioral_analysisAggregate patterns, momentum and risk across a set of habitshabit_analysisPer-habit completion, streaks and consistencybehavioral_insightsCondensed highlights and watchoutshabit_recommendationRanked next-best actions with rationale and priorityhabit_planningStructured multi-week plan with weekly targets and milestonesreliability_analysisReliability scoring for recurring tasks, pipelines, and jobs| Endpoint | Per call |
|---|---|
POST /agent/v1/habits/analyzeAggregate behavioural analysis, up to 25 habits per call | $0.05 |
POST /agent/v1/habits/planStructured multi-week habit plan | $0.10 |
POST /agent/v1/habits/insightsCondensed highlights and watchouts | $0.05 |
POST /agent/v1/habits/recommendRanked next-best actions | $0.03 |
POST /agent/v1/reliability/analyzeOperational task and pipeline reliability scoring | $0.08 |
Price is per request, not per habit — batching up to 25 habits costs the same as one. Paid in USDC on Base via the x402 protocol.
Volume rebate: Verified spend earns credits back: 5% from $10 lifetime, 10% from $100, 15% from $1,000. Full terms →
Call the endpoint. It answers 402 with the payment terms in the
PAYMENT-REQUIRED header. Sign them with your wallet and retry with
X-PAYMENT.
# 1. Probe or ask — returns 402 with the payment terms curl -X POST https://api.mydev-tools.tech/agent/v1/habits/analyze \ -H 'Content-Type: application/json' \ -d '{"habits":[{"name":"Morning run","completion_rate":0.72,"current_streak":9,"best_streak":21}]}' # 2. Sign the challenge with your Base wallet, retry with X-PAYMENT header
stateless agent callsAgent endpoints operate strictly on the payload supplied. No stored personal data is returned.personal baseline isolationYour baseline is computed exclusively from your historical data. No cross-user profiling or stereotype comparisons.minimised AI exposureOnly derived statistical features reach external language models. Raw notes stay in local storage.complete cascade deletionAccount deletion permanently scrubs all habits, logs, vectors, baselines, and models.USDC on Base x402 v2 ABPE Engine Active OpenAPI 3.1
/healthLive service status/openapi.jsonMachine-readable contract (includes /v1/behavior/*)/v1/capabilitiesCapability manifest with schemas and prices/llms.txtPlain-text summary for LLM crawlers