ac2185b8...8835dfQuant trades the chart. Catalyst trades the story.
Catalyst is a risk-governed agent desk. Specialist scouts read Bitget data and Agent Hub signals, Qwen writes the thesis, and the governor decides: trade, or NO TRADE until the evidence clears risk. No trade is still a decision.
Forward ledger
6/6 refused trades crossed their stop band.
Each cron receipt is committed to GitHub, then the ledger aggregates its shadow-audit row: the approved trade path versus the forced-thesis baseline Catalyst refused. “NO TRADE” becomes measurable instead of cosmetic.
+0.7422% shadow-audit discipline baseline across 34 cycles; 6/6 vetoed forced setups crossed the stop band.
da9ae58a...2421fa1cd842fe...1585fafab238a3...801a393b9496aa...6d6942Risk-governed agent desk
Five scouts debate before the governor decides.
Catalyst is no longer just a single model call. Each receipt shows the specialist scouts, Qwen's synthesis, source status, consensus, and the deterministic risk governor that can veto the setup.
Agent Hub technical-analysis verdict BEARISH; RSI 44.30; MACD death_cross; MA trend mixed; ATR 1.77%
realized volatility proxy 2.12%, regime range
Agent Hub macro context received: macro_indicators count: 3; macro_indicators title: Minutes of the Board's discount rate meeting on April 2…; macro_indicators link: https://www.federal...
Agent Hub news returned 8 market headline(s); 4 matched BTC: Bitcoin price slips toward $62K local lows as bear-market history keeps repeating | Bitcoin bottom? These 4 charts hint at BTC p...
Funding 0.0033%, open interest 31.63K contracts; Bitget derivatives positioning is balanced
Technical indicators converge on downside with a MACD death cross, RSI below 50, and negative short-term momentum. Macro risk-off conditions persist with VIX elevated near 21.6 and recent F...
Protection ledger
Standing down has a receipt too.
Most trading demos only celebrate fills. Catalyst also records the trade it refused to force, then runs a labeled shadow simulation so judges can see what the governor protected against or deliberately ignored.
What the governor allowed
Reviewed SHORT BTCUSDT
The setup cleared the governor; this shadow row mirrors the approved reference trade.
Simulation only. This is not a real order and not a PnL claim; it is the governor's audit trail.
Shadow outcome
-1.80%
ac2185b8acd69b84c4a7269fc0df7e47de4c2ee51ef087356a3933ad888835df
Latest autonomous cycle
Catalyst entered; outcome pending.
live Bitget + Qwen receiptTechnical indicators converge on downside with a MACD death cross, RSI below 50, and negative short-term momentum. Macro risk-off conditions persist with VIX elevated near 21.6 and recent FOMC minutes, while news highlights market capitulation and historical cycle models projecting a drop toward $50K–$53K. In a ranging market, this confluence favors short exposure near resistance.
Risk governor
All deterministic risk checks passed.
- Passlive signal coverage: 6 live / 3 required
- Passconviction floor: 0.65
- Passvolatility cap: 0.02467
- Passmomentum not extreme: -0.01456
ac2185b8acd69b84c4a7269fc0df7e47de4c2ee51ef087356a3933ad888835df
Historical trade-cleared replay
Catalyst entered the trade.
real Bitget candles + GetAgent proof, not the latest live cycleHistorical replay: GetAgent run pbrun-897d2346352e showed a disciplined LONG setup while Catalyst had 3/3 real replay sources. The same deterministic governor cleared a bounded position instead of forcing full exposure.
Risk governor
All deterministic risk checks passed.
- Passlive signal coverage: 3 live / 3 required
- Passconviction floor: 0.7961
- Passvolatility cap: 0.01743
- Passmomentum not extreme: -0.00147
dc2a03be82792da3dcb1c3e40f63ecb46678bf0597bc58de389340cb04a93f8a
Browser verifier
Change one value and the receipt fails.
Catalyst receipts are checksum-backed. This browser recomputes the public checksum from the JSON artifact, then lets judges tamper with the decision to prove the receipt is not a decorative card.
Receipt checksum is live-verifiable.
Receipt untouched. Browser hash matches the public checksum.
BTCUSDT proof
APPROVEac2185b8...8835dfac2185b8...8835dfac2185b8...8835dfopen JSONPerception layer
The agent shows its inputs before it acts.
The current live receipt uses Bitget candles and Bitget derivatives positioning. Agent Hub technical-analysis is included when that endpoint returns usable data; macro and news use Agent Hub first, then labeled cache or fallback context that does not count as independent convergence.
24h momentum -0.88%, 72h momentum 0.03%
bearishrealized volatility proxy 2.12%, regime range
unclearAgent Hub technical-analysis verdict BEARISH; RSI 44.30; MACD death_cross; MA trend mixed; ATR 1.77%
bearishAgent Hub macro context received: macro_indicators count: 3; macro_indicators title: Minutes of the Board's discount rate meeting on April 2…; macro_indicators link: https://www.federalreserve.gov/newsevents/pressreleases/mon…; macro_indicators pubDate: Tue, 26 May 2026 18:00:00 GMT
risk_offAgent Hub news returned 8 market headline(s); 4 matched BTC: Bitcoin price slips toward $62K local lows as bear-market history keeps repeating | Bitcoin bottom? These 4 charts hint at BTC price dropping to $50K | Bitcoin 'normal' 4-year cycle puts focus on $53K low before 2028 BTC price high
catalyst_watchFunding 0.0033%, open interest 31.63K contracts; Bitget derivatives positioning is balanced
neutralDecision pair
One agent, two honest outcomes.
The latest live cycle cleared risk and entered a paper trade; the forward ledger shows the trades the governor refused and how the market later graded them. Catalyst is not a no-trade poster; it is an autonomous decision desk.
1 / Live current cycle
Risk cleared, so Catalyst entered.
The current cycle selected BTCUSDT, used live Bitget market data and Qwen, then ended in APPROVE. The paper fill is recorded while the live outcome remains pending.
ac2185b8...8835dfreceipt JSON2 / Vindicated veto ledger
6/6 refused trades crossed their stop band.
Example: Catalyst refused BTCUSDT short; the forced setup later crossed its 1.80% stop band.
da9ae58a...2421faledger JSON3 / Sponsor proof
Agent Hub, Qwen, GetAgent, checksum.
Live receipt stack includes technical-analysis: bearish and qwen3.6-plus. GetAgent run pbrun-e56438803922 records Sharpe 2.564. The decision-pair verifier checks the live receipt and trade-cleared replay.
npm run verify:decision-pairplaybook JSONAutonomous cycle history
Catalyst now leaves a receipt trail.
Each cycle scans live Bitget markets, uses live positioning when available, asks Qwen for the strongest candidate, then lets the deterministic governor approve or reject before writing a checksum-backed receipt.
ac2185b8...8835dfda9ae58a...2421fa1cd842fe...1585fafab238a3...801a393b9496aa...6d6942bbed0414...c0335bLive market selection scan
Catalyst chooses before it reasons.
The agent scans live Bitget markets first, ranks each setup deterministically, then sends only the strongest candidate to Qwen and the risk governor. This prevents a human from hand-picking the receipt.
GetAgent Playbook proof
Published on GetAgent Studio.
Catalyst Discipline is listed in Studio after a cloud backtest. Studio card metrics are shown separately from the raw API run because Studio normalizes reports for ranking.
Reproducible proof
Same receipt, same checksum.
Judges can replay the local loop and compare the checksum with the public receipt. The current live cycle can approve or block depending on the risk governor.
Current receipt checksum
ac2185b8...8835df
The full checksum is stored in public/demo-data/latest-cycle.json.
npm test npm run forward:ledger npm run playbook:summary -- --write-public npm run cycle:strict npm run build:static