Autonomous Loop

Fload's Autonomous Loop is the engine that ties everything together. It's a six-system architecture that enables Fload to observe, decide, act, and learn — continuously and without manual intervention.

This page explains how the systems work together. Each system builds on the ones before it.

System 1: Autonomous loop

The core execution cycle. Fload runs on a configurable schedule (2–6 times daily) and follows an observe-review-decide-dispatch loop.

Draft mode vs Agent mode

• Draft mode — Fload generates recommendations and queues them as pending actions. You review and approve before anything happens. This is the default.
• Agent mode — Fload executes decisions autonomously within safety guardrails. It still logs everything, but moves without waiting for approval.

You can switch modes at any time in Settings → Orchestrator. Most teams start in draft mode and graduate to agent mode once they trust the system.

System 2: Orchestrator dispatch

The Orchestrator is the brain of the loop. Each cycle, it:

1. Observes — collects results from all agent runs and pulls the latest metrics
2. Reviews — identifies cross-agent patterns (e.g., negative reviews + retention drop = possible bug)
3. Decides — prioritizes what needs attention based on impact and urgency
4. Dispatches — creates pending actions or auto-executes them, depending on your mode

The Orchestrator also handles event-triggered mini-cycles — reacting to anomalies, review surges, ad spend issues, or submission rejections in real time.

System 3: Memory system

Fload remembers past decisions and their outcomes. This is what makes it get smarter over time.

• After every action, the system schedules a check-back to evaluate the result
• Positive outcomes reinforce the decision pattern ("pausing that campaign improved ROAS by 15%")
• Negative outcomes are flagged so the same mistake isn't repeated
• Memory is per-app, so recommendations are tailored to each app's unique patterns

The memory system powers the Orchestrator's decision-making. Without it, every cycle would start from scratch.

System 4: Agent write actions

Agents don't just analyze — they act. Write actions are coordinated across agents:

All write actions flow through a unified pipeline:

• In draft mode, they appear as pending actions for your approval
• In agent mode, they execute automatically and you're notified
• Every action is logged with before/after values for easy rollback

Safety guardrails (budget caps, daily action limits, rollback logging) apply regardless of mode.

System 5: Feedback loops

The loop isn't just circular — it improves over time through structured feedback:

• Orchestrator feedback — tracks which recommendations you approve vs. dismiss, and adjusts future priorities
• Review Agent feedback — learns from your edits to draft replies, adapting tone and content
• Growth Audit feedback — compares scores across runs to measure whether recommendations are working
• Ad performance feedback — measures the outcome of budget and bid changes to inform future optimizations

The more you interact with Fload, the more accurately it reflects your preferences and judgment.

System 6: Frontend agent integration

Every system above surfaces in your Fload dashboard:

• Daily briefings — delivered via email, Slack, or Discord with key metrics, agent activity, and recommended actions
• Pending actions — a unified queue of everything waiting for your approval, with one-click approve or bulk actions
• Activity log — a complete timeline of every action taken, by whom (you or the agent), and the outcome
• Agent status — see which agents ran, when, and what they found
• Memory insights — view what the system has learned about each app

The dashboard is designed so you can stay informed in 5 minutes or go deep when you need to.

How it all fits together

Schedule / Event trigger
        |
   [System 1] Autonomous Loop (draft or agent mode)
        |
   [System 2] Orchestrator observes → reviews → decides → dispatches
        |
   [System 4] Agent write actions execute (with guardrails)
        |
   [System 3] Memory records the decision and outcome
        |
   [System 5] Feedback loops adjust future behavior
        |
   [System 6] Frontend surfaces everything to you

Each cycle feeds the next. Actions produce outcomes. Outcomes update memory. Memory improves decisions. Decisions produce better actions.

Getting started

The autonomous loop runs automatically once you've:

1. Connected at least one data source (App Store Connect, Google Play, etc.)
2. Activated agents — use Smart Setup to auto-configure Review, Growth Audit, ASO, and Anomaly Detection agents with one click, or configure agents manually
3. Set your Orchestrator preferences (run frequency, approval mode, notification channel)

All agent pages are always visible — there are no activation gates. Your data and dashboards are free. AI actions (Orchestrator cycles, agent dispatch, write actions) require a Pro subscription ($50/month, 14-day free trial).

Start in draft mode to see what Fload would do. Switch to agent mode when you're ready to let it act on your behalf.

See Orchestrator for detailed configuration options.