Indicator Platform Guide

How Indicator HQ Works

HQ is Indicator's autonomous agent layer — four AI agents that study, scan, and act on markets continuously while you focus on decisions.

Overview

4 Founder Agents

Robin · Sherlock · Jarvis · Stark

Self-Learning

Studies GitHub repos to acquire skills

Always On

Autopilot runs every 15 min

Think of HQ as a small autonomous research team that never sleeps. You give them a mission (or let them run on autopilot), and they coordinate to scan the market, analyze tickers, learn new techniques, and alert you when something interesting happens. Every action earns XP, every level-up evolves the agent, and high-level agents can spawn specialists — your team grows over time.

Meet the Agents

RobinMarket Pulse

Reads the market's mood. Scans bellwether stocks, sentiment signals, and overall direction.

Market BEARISH: PTT −3.52% · ADVANC +0.27% · CPALL 0% · SCC −1.47% · KBANK −0.78%

SherlockIndustry Scan

Hunts down sector trends. Scans 23+ industries and surfaces which sectors are hot or cold.

Scanned 23 sectors. Top: Petrochemicals +104 · Electronic Components +95 · Industrial Materials +73

JarvisSignal Monitor

Watches your tickers 24/7. Fires alerts on RSI extremes, price breaks, and volume spikes.

Checked 10 tickers. Alerts: COP RSI overbought (70.12) · COP +1.67%

StarkTechnical Dive

Deep technical analysis. RSI, MACD, Bollinger bands, crossovers — finds the best setups.

Top pick: BDMS. MACD_BEARISH on DELTA · MACD_BULLISH on ADVANC · KBANK MACD_CROSSOVER

Capabilities

Every subtask requires a specific capability. The router assigns it to the agent best matched:

Market Pulse

Overall sentiment from bellwether stocks (PTT, ADVANC, CPALL, SCC, KBANK)

Industry Scan

Scores 23 sectors on momentum, flags rotation winners

Signal Monitor

RSI, price, volume alerts on a watchlist of tickers

Technical Dive

MACD, Bollinger, crossovers, trend strength on priority picks

Research

Self-study mode — learns new skills from GitHub repos on chosen topics

Evolution Ladder

Agents earn XP from every action. Level up → evolve tier. At Lv 5 and Lv 10 a founder spawns a born agent (specialist child).

Spark

Lv 1–4

Electric

Lv 5–9

Neural

Lv 10–14

Quantum

Lv 15–19

Cosmic

Lv 20+

Spark— Newborn agent, learning the ropes

Electric— Can spawn first born agent

Neural— Spawns second born agent

Quantum— Deep specialization unlocked

Cosmic— Peak evolution — masterful

Born Agents

When a founder levels up, they spawn children: Robin → Oracle (Lv5), Cipher (Lv10) · Sherlock → Quant, Scout · Jarvis → Sage, Weaver · Stark → Forge, Pixel. Born agents start fresh but inherit their parent's specialty focus.

How Missions Work

A mission is a natural-language goal. HQ plans it, assigns agents, and executes — all autonomously.

You state a mission

"Scan Korean semi stocks and flag any MACD crossovers"

AI Planner decomposes

GPT-4o-mini breaks it into subtasks with dependencies (DAG)

Router assigns agents

Each subtask routed to the agent with matching capability & highest XP

Agents execute in order

DAG-aware — dependent tasks wait for upstream to finish

Findings broadcast

Each agent shares results on the message bus; failed tasks recover or skip

Mission resolved

Aggregated summary + per-agent contributions saved to archive

Recovery & Resilience

If a subtask fails, the Recovery Policy Engine classifies the failure and decides: retry, fallback to a different capability, or skip. Dependent downstream tasks are auto-skipped if their deps fail.

Autopilot

Turn on autopilot and HQ runs every 15 minutes — even when you're not watching.

Heartbeat

Every 15 min, all 4 founders execute their primary capability

Decision Engine

Prioritizes pending missions and message inbox before scheduled work

Interest Drift

Free-will learning: agents wander based on curiosity style (explorer / balanced / specialist)

Evolution Check

After each heartbeat: check level-ups, tier upgrades, and birth conditions

Quick Start

Open HQ

Navigate to the HQ tab — see all 4 agents with their current level, tier, and focus.

Run All once

Click "Run All" in the top-right. This triggers a manual heartbeat — every agent does their job once. Check the activity feed for results.

Start Autopilot

Click "Start Autopilot" to schedule automatic runs every 15 min. Agents now work in the background — revisit anytime to see what they found.

Go to HQ