A small platform,
deliberately.

A CLI and MCP server statically read your routes, analytics calls, and product journeys, then draft scored events, multipliers, and levels mapped to what your app actually does.

The blank page is why gamification stalls. Deciding what to score is harder than calling the API — so the agent does it from your real code, and you approve a diff.

Your whole economy lives in a versioned config file in your repo — reviewed in pull requests, deployed through CI like any other code.

Config trapped in a vendor dashboard can't be reviewed, diffed, branched, or rolled back with your app. Yours travels with your codebase.

Replay a synthetic user journey through your config in CI and assert outcomes — “after 7 lessons this user is level 3 with a 5-day streak.”

Economies drift and silently break on a rebalance. If scoring is code, it should be testable code — guarded in your pipeline like everything else.

Scores are a projection over an immutable event log. Each change is a named version; recalculate forward, from a cutoff, or in full — with audit trail and one-click rollback.

You will change multipliers as you learn. Doing it destructively rewrites history with no record; doing it safely needs replay from raw events.

Pipe live production events into a draft scoring version without touching real scores. Watch how the economy would behave, then flip it when it's right.

Changing a points economy is genuinely risky — you can demoralize loyal users or open a farming loop. Shadow mode turns a leap into a measured decision.

Every scored event returns its full breakdown — base × multiplier × each bonus, plus any caps or cooldowns that applied. Expose it as a “why these points?” trace.

Arbitrary-feeling points erode trust and are impossible for support to debug. A transparent computation is both fairer and answerable.

An agent watches aggregate outcomes — cohort retention, streak survival, score distribution — and proposes a scoring-version diff. You review and approve, just like setup.

Tuning an economy is expert work most teams do once and forget. An advisor that suggests data-driven rebalances keeps the loop healthy over time.

Streakline emits domain events — level.up, achievement.unlocked, streak.broken — as webhooks. You wire them to the messaging stack you already run.

Gamification and notifications are different crafts. Doing one well and emitting clean events beats bundling a mediocre mailer you can't replace.

Stream your raw events and scores to your own warehouse (BigQuery/Snowflake) or pull the full history via API and webhooks — your data stays portable even though we host it.

Behavioral data is most useful next to the rest of your data, and teams fear lock-in. Continuous export means you can JOIN it in your warehouse and leave anytime.

Because rules are declarative, the same engine ships to your client and computes an optimistic score instantly, then reconciles with the server of record.

A points animation that waits on a network round-trip feels broken — especially on mobile. Instant, correct-by-construction feedback is the difference.