SigmaVue overview
The operator's manual. SigmaVue is twelve specialist AI agents managing your prop firm accounts under your rules — while you're at your day job, your kids' game, or asleep. This page is the single source of truth for how the agent floor works, end to end, in plain English.
~/.claude/skills/, and prompts for your token.
curl -fsSL https://install.sigmavue.io | sh
How it works
You connect your prop firm accounts (Apex, TopStep, TPT, Leeloo — 18+ firms supported), set your daily base hit and risk profile, and the agents take over. Sigma orchestrates, Carter scans, Annalisa reads news, Bill sizes for the base hit, Wags vetoes anything that breaks compliance, Wendy & Devin execute, Bonnie journals.
No charts to watch. No webhooks to wire up. No VPS to pay for. Everything runs on the SigmaVue platform — you watch the floor, or close the tab.
5-minute setup
Three clicks, three minutes each. We OAuth into your prop firm — we never see your password — and route automatically to Tradovate, ProjectX, or TradeLocker based on your firm and account state.
40/35/25 across three Russian Doll layers — most cashes early, the rest runs to full TP. Wags pauses the day when the inner layer locks.Example: a single trade lifecycle
Here's the audit trail of one ES.M26 long signal — start to fill — that ran on April 22, 2026 at 10:42:18 ET. End-to-end took 3 seconds.
{ "trade_id": "ES_M26_1745324538", "symbol": "ES.M26", "entered_at": "2026-04-22T14:42:21Z", "chain": [ { "t": "10:42:18", "agent": "Carter", "action": "signal_long", "reasoning": "VWAP stretch 1.7σ + Maff trend day 0.71 + Bill headroom OK" }, { "t": "10:42:19", "agent": "Bill", "action": "sized", "qty": 2, "reasoning": "Inner base hit at +6 ticks · 2 ct = $25/tick" }, { "t": "10:42:20", "agent": "Wags", "action": "compliance_pass", "reasoning": "Apex rules clear · within session · EIA in 90min noted" }, { "t": "10:42:21", "agent": "Wendy", "action": "order_filled", "limit": 5871.75, "tp": 5876.25, "sl": 5869.75, "layers": "russian_doll" } ], "layers": { "inner": { "qty": 0.40, "target": "base_hit" }, "middle": { "qty": 0.35, "target": "tp_70pct" }, "outer": { "qty": 0.25, "target": "runner" } } }
Every field above is exposed in the audit log export and on the Sigma Decisions screen inside the app. chain is immutable and timestamped at agent decision-time, not server-receive-time, so post-trade analysis stays honest.
The Russian Doll exit
Most retail trades give up >60% of MFE on the round trip. The Russian Doll splits each position into three layers so the inner layer locks fast, and the outer layer runs only when the trend confirms.
| Layer | Size | Target | Hit rate (2y backtest) |
|---|---|---|---|
| Inner | 40% | Base hit (~6 ticks ES) | 81% |
| Middle | 35% | 70% of TP | 49% |
| Outer | 25% | Runner / structural target | 28% |
The 12 agents
Twelve specialists, one orchestra. Each agent owns a single decision surface — they argue, defer, and pass the trade through a hand-off chain that's logged in real time. Below: who they are, what they own, and what triggers them. No prompts, no internal scoring math — that's the moat. But every operator should know the names on the floor.
Routes every signal through the council, weighs disagreement, and decides which agent has the floor on a given trade. Sigma never trades alone — but no trade leaves the building without Sigma's stamp.
Generates the long/short signals. Reads structure, momentum, and stretch. Argues hardest in the council when conviction is high — and stays quiet when it isn't. Most floor time of any agent.
Reads the wire — econ data, FOMC, earnings, geopolitics, and prop-firm-relevant macro events. Vetoes Carter when an EIA print is 3 minutes out. Keeps Bonnie's journal grounded in real-world context.
The skeptic. Every Carter signal passes through Ben for quality scoring before it's sized. If the regime, conviction, or context doesn't pass Ben's bar, the signal dies quietly — no broker call.
When you want Sigma to ask first, Lion is the one who asks. Configurable: full auto, ask-on-large, ask-on-novel, or ask-on-everything. Pushes to your phone, waits as long as your config allows, then defers to Wags' standing rules.
Translates a signal into how many contracts. Sized for the inner Russian Doll layer at base hit, never above account headroom, never above the daily loss limit, never above your active risk profile.
The veto. Knows every prop firm's rules — daily loss, drawdown trail, news lockouts, position scaling, weekend exposure, EOD flatten — and refuses any trade that breaks them. Wags is the most-blocked, most-loved agent on the floor.
Compresses the council's debate into a single decision packet. Ties Carter's signal, Ben's filter, Annalisa's macro, Bill's size, and Wags' compliance into one row in the audit log — with the reasoning of each.
Picks the broker. Splits the order across funded accounts when you have multiple. Handles Tradovate / ProjectX / TradeLocker routing automatically based on firm and account state. Retries on transient failures, never on rejected.
Watches every open position across every funded account in real time. Computes layer-by-layer realized vs. unrealized, fund-level base-hit progress, and triggers Wags' "pause the day" when the inner layer locks across all accounts.
Writes the day's recap. End-of-session summary, trade-by-trade reasoning, regime calls, what got vetoed and why. The piece you'd send your trading partner — except it writes itself, and it's grounded in the audit chain.
The friendly one. Walks new operators through setup, answers in-app questions, surfaces relevant docs in chat. If something feels broken, talk to Donna first — she handles the triage and pulls in a human if needed.
Supported prop firms
18 firms supported in v3. Wags ships with a per-firm rule profile — daily loss, trailing drawdown, news lockouts, position scaling, EOD flatten, weekend exposure — and updates automatically when a firm changes its eval or live rules. If your firm isn't here, ask in Discord and we'll evaluate.
MCP server Operator+
The headline integration. Connect SigmaVue to Claude Code, Codex, Cursor, or any MCP-aware client and ask questions about your floor in plain English. The MCP server is hosted at mcp.sigmavue.io and authenticates with a per-account token — you don't run anything locally.
Three lines of config in your ~/.claude/settings.json and you're live:
{
"mcpServers": {
"sigmavue": {
"url": "https://mcp.sigmavue.io",
"headers": { "Authorization": "Bearer $SIGMAVUE_TOKEN" }
}
}
}
Then in Claude Code, just talk to your floor:
# Read your floor claude "show me yesterday's vetoes and what they cost" claude "summarize Carter's win rate this week vs last week" claude "which prop firm rules am I closest to violating right now?" # Trader+ tier — write actions (governed) claude "pause trading on apex-1247 for the rest of the day" claude "set my risk profile to conservative until further notice"
Tool catalog
| Tool | Tier | Description |
|---|---|---|
| sigmavue.audit.list | Operator | Last N decisions across accounts. Filter by agent, by firm, by outcome. |
| sigmavue.audit.get | Operator | Full hand-off chain for a single trade ID — every agent's reasoning row. |
| sigmavue.positions.list | Operator | Open positions, by account, with layer-by-layer realized vs. unrealized. |
| sigmavue.pnl.summary | Operator | Aggregated PnL by account, day, week, month. Includes base-hit attribution. |
| sigmavue.config.get | Operator | Risk profile, base hit, account list, agent toggles. Read-only. |
| sigmavue.alerts.subscribe | Operator | Server-sent stream of fills, vetoes, and base-hit lock events. |
| sigmavue.config.set_risk_profile | Trader+ | Switch between Conservative · Balanced · Aggressive (named profiles only). |
| sigmavue.trading.pause | Trader+ | Pause one or all accounts. Open positions follow Russian Doll exits. |
| sigmavue.trading.resume | Trader+ | Resume from pause. Wags re-validates rules before any new entry. |
Claude Code skill Free download
One-click install for Claude Code users. Adds canned prompts, common workflows ("morning briefing", "post-market journal", "rule-violation check"), and shorthand commands. Works on top of the MCP server above.
~/.claude/skills/. Or run the one-line install in your terminal.curl -fsSL https://docs.sigmavue.io/sigmavue.skill.md \
-o ~/.claude/skills/sigmavue.skill.md
export SIGMAVUE_TOKEN="sk_...". Tokens are scoped to a single account by default.mcp.sigmavue.io URL in their respective settings — the tool catalog is identical across clients.Audit log · CSV export
Every decision is exportable to CSV directly from the dashboard — no API needed. Settings → Audit → Export. Filter by date range, agent, firm, or outcome. The CSV is immutable: timestamp at agent decision-time, full reasoning chain, layer attribution, and final fill.
Daily Base Hit
The single number you set on day one — and the number every agent on the floor optimizes for. Pick a dollar amount you'd be happy to take home each day. SigmaVue splits it 40 / 35 / 25 across three Russian Doll layers: most cashes early at the inner layer, the middle layer banks at 70% of TP, the outer layer runs.
Per-account targets compound at the fund level. $150 × 4 funded accounts = $600/day fund-level base hit. Wags watches the cumulative number across accounts and pauses the day automatically once the inner layer locks across all of them — no over-trading, no give-back, no "one more setup."
$150 is a great day for you, pick $150. The system's edge is consistency — over a year, locking the base hit on 80% of trading days beats swinging for a 5× target on 30% of them.Risk profiles
Three named profiles — Conservative · Balanced · Aggressive — set how much of your account headroom Bill is allowed to use per trade. No custom sliders. Profiles are intentionally coarse so the floor never accidentally drifts to a config you didn't think through.
| Profile | Per-trade DD | Day cap | Best for |
|---|---|---|---|
| Conservative | 1.5% | 3.5% | Combine / Eval phases · drawdown-strict prop firms |
| Balanced | 2.5% | 5.0% | Funded accounts · default for most operators |
| Aggressive | 4.0% | 7.0% | Funded accounts only · high-conviction operators |
Wags rejects Aggressive on Eval-phase accounts regardless of your settings — the trailing-DD math on most prop firms makes Aggressive sizing a first-week-blowup risk during evaluation. Pass the Eval first, opt in second.
Compliance veto
Wags is the agent who says no. Every signal — every signal — passes through Wags before it reaches the broker. Wags' rule engine is loaded with a per-firm profile: daily loss limit, trailing drawdown, news lockouts, position scaling, weekend exposure, EOD flatten. If any rule would be broken, Wags vetoes.
Common veto reasons (and what they mean):
news_lockout— High-impact econ data is within the firm's lockout window. Resumes automatically when the window expires.dd_proximity— Account is within 30% of trailing drawdown. Bill could size for it; Wags refuses.weekend_exposure— Weekend hold not permitted by the firm. Trade flattens at EOD Friday.position_scaling— Firm prohibits scaling beyond initial size. Council wanted to add; Wags blocked.session_window— Outside the firm's allowed trading session for that contract.
Emergency halt
Top-right of the dashboard, always one click away. The big red button. Halt does three things in this order:
- Cancels all working orders across every connected account.
- Flattens all open positions at market — Russian Doll layered exits are skipped.
- Pauses every agent from generating new signals until you manually resume.
Resume is manual and per-agent — when you click resume, Wags re-validates rules and Bill re-checks headroom before any new entry. The full halt + resume sequence is logged with timestamps and operator confirmation. Use halt sparingly; the agents recover gracefully from one halt, but two halts in a week tells the team to check in with you.
Your first week
The honest playbook for the first 7 days on the floor. Don't override the agents. Don't over-trade. Don't sit and watch every tick. Three rules.
$50). Watch the floor trade for two sessions. Read the audit log. No live capital yet. The goal is to see what Carter signals, what Wags vetoes, and how Bonnie writes the journal — not to make money.$100. Let it run. Don't override. Read Bonnie's EOD journal each evening — that's where you'll see whether the floor is reading the market the same way you would.System status
Real-time status of every component the agents depend on. Public status is at status.sigmavue.io. Below is the live snapshot — refreshes every 30 seconds.
Glossary
The vocabulary the floor uses, in plain English. If a term in the dashboard isn't here, ask in Discord and we'll add it.
Changelog
Public release notes. Internal pre-release builds are not listed. Every release lands first on staging for 48 hours before promotion.
- MCP server (Operator+): connect SigmaVue to Claude Code, Codex, or Cursor. 9-tool catalog covering audit, positions, P&L, config, alerts.
- Claude Code skill: free download with canned workflows (morning briefing, post-market journal, rule-violation check).
- Tradeify added as supported prop firm. Bringing total to 18.
- Russian Doll inner-layer hit rate improved
78% → 81%after Bill's sizing recalibration. - Wags rule-profile auto-update for Apex's new EOD requirement (April 4 firm announcement).
- Bonnie auto-journal now uses operator's own voice profile when configured.
- Strategy catalog deleted. The agents pick now. Operators don't choose strategies — they set base hit + risk profile.
- Office screen redesigned as the trading floor. 3×3 agent grid + orchestrator chip strip.
- 10 fresh PRDs replacing the previous 47. Methodology refresh.
- Founding 100 program launch.
23 / 98spots remaining at$4,200/yr locked-for-life. - Community feed and live stream moved into the app. No more separate Discord trading rooms required.
What to read next
- Set up the MCP server — connect SigmaVue to Claude Code, Codex, or Cursor in three lines of config.
- Install the Claude Code skill — canned workflows, morning briefing, post-market journal, rule-violation check.
- Meet the 12 agents — who's on the floor, what they own, and what triggers them.
- Pick your prop firm — 18 supported, with auto-applied rule profiles.