// DEMOS

See the systems in motion

Two live systems you can open right now, and four capability demos that run right here in your browser.

LIVE SYSTEMSReal products, running now. Explore how they work here, then open the real thing.
LIVE · MULTI-AGENT SYSTEM

Agent Office

A simulated trading firm staffed by AI agents. Seven departments research, build, backtest, trade and police strategies on real exchange prices with paper money, and a council of voters has to approve every entry before it opens.

  1. Agents message each other for real: every white bubble in the office is an actual agent message
  2. Nothing trades until a candidate clears the backtest gates and the council votes the entry through
  3. Every signal, vote and trade lands in a hash-chained ledger that can be verified end to end
SNAPSHOTsimulation only · paper money · no real orders
HOW IT REALLY WORKS
Ledger0 events this cycleappend-only
ResearchReads the market so the other rooms do not have to guess.
  • Research LeadMODEL-DRIVEN AGENTLeads research and assigns studies
  • News ScoutAUTOMATED JOBSearches the web for news on the traded coins and flags breaking stories
  • Sentiment AnalystAUTOMATED JOBScores crypto headlines bullish, bearish or neutral
  • Chan AnalystAUTOMATED JOBReads the 15m Chan structure: pivots, strokes, divergence
  • Derivatives AnalystAUTOMATED JOBReads funding, the futures curve, implied volatility and options positioning
OFFICE FEED · READYscripted replay · modelled on real office traffic
Pick a cycle to watch the departments hand work to each other.
LIVE · AI LEAD DESK

Kaunter, the AI front desk

An AI front counter for shops that sell on chat. It reads the comments under your posts, finds the people who are actually asking, messages them in their own language, quotes from your price list, books the visit, and calls a person in when it should.

  1. Scan a post: every comment is classified and scored by intent and recency
  2. Approve the drafted first message, written in the commenter’s language
  3. Play the customer: get a quote from the price grid, then book or push back
01 SIGNALS02 FIRST MESSAGE03 CHAT
Kelab Trade-In Phone Malaysiademo post

Traded in my iPhone 15 Pro Max 512GB today, RM3,100 total 🎉

score = intent × recency (24h half-life)
Demo shop: prices, stores and calendar are invented. Quotes come only from the price grid.
CAPABILITY DEMOSOne per capability, simulated in your browser. No backend involved.
AI_CORE · CONVERSATIONAL

Conversational AI copilot

A branded copilot grounded in your knowledge base. Ask it anything — it understands intent and responds in natural language.

  1. Type a question or tap a suggested prompt
  2. The copilot matches intent against its knowledge base
  3. It streams back a grounded, on-brand answer
Hi! I'm SpaceHero's assistant. Ask me anything about what we build, timelines, or support.
Hi! I'm SpaceHero's assistant. Ask me anything about what we build, timelines, or support.
AI_CORE · ANALYTICS

Real-time analytics

Streaming pipelines turn raw events into live business signal. Watch the metrics update in real time and reshape across time ranges.

  1. Events stream in and update KPIs live
  2. Pick a time range to reshape the series
  3. Pause the feed to inspect a moment
LIVE
Revenue
RM 48.2k
▲ 2.1%
Active Users
1.2k
▲ 1.2%
Conversion
3.8%
▲ 0.3%
Latency
42ms
▲ 4%
BOT_NET · AUTOMATION

Automation flow runner

Background agents triage, route, escalate, and report — so your team stays in flow. Run the pipeline and watch each step execute.

  1. Press Run to ingest an incoming event
  2. The agent classifies, routes, and escalates
  3. A summary report is dispatched automatically
📥
Ingest
→
🧭
Triage
→
🔀
Route
→
⚠️
Escalate
→
📊
Report
Press “Run automation” to watch the agent pipeline execute.
NEURAL · MACHINE LEARNING

Intent classifier

A trained model scores incoming messages across intent classes in milliseconds. Type a message and see the confidence breakdown.

  1. Enter a message or pick an example
  2. The model scores every intent class
  3. The top prediction wins with a confidence score