Quick answers about AlphaFlowSeven's AI council, portfolio simulation, pricing, and the platform's data infrastructure.
Yes. New users get up to three AI councils on a curated set of tickers — completely free. No credit card required. You only pay if you scale beyond the free tier, and even then billing is usage-based rather than a subscription.
No. AF7 is a signal platform with a built-in portfolio simulation. Every signal can be routed into a simulated position so you can evaluate the council's calls without committing real capital. There is no on-exchange execution and no exchange account is required.
Each of the six agents in a council can run on a different model. Seven providers are supported: Anthropic Claude (Fable 5, Sonnet 5, Opus, Haiku), OpenAI (the GPT-5 family), Google Gemini (3.5 Flash, 3.1 Pro, 3.1 Flash-Lite), xAI (Grok 4.1 Fast), DeepSeek (V4 Pro and Flash), Moonshot AI (Kimi K2.6), and Zhipu (GLM).
Yes. Every one of the six seats can run a different model from any of the seven supported providers (Claude, OpenAI, Gemini, Grok, DeepSeek, Kimi, GLM), and each agent's system prompt is fully editable — per council — with a prompt library to save and reuse setups. The Council Evaluator audits a council's recent calls against their actual outcomes and proposes concrete prompt rewrites for each agent, and warns you if a live prompt has drifted from the version that was evaluated. The three-tier, six-seat structure itself is fixed — that adversarial shape is the product — so AF7 doesn't offer arbitrary custom agents, data uploads, or user-defined flows; it gives you deep control over the models and prompts inside a proven structure.
Three reasoning tiers. Tier 1 — Technical, Momentum, and Risk analysts — read the market in parallel. Tier 2 — Bull and Bear advocates — build the strongest case for each direction from Tier 1 findings. Tier 3 — the Lead Trader — synthesises everything plus its own last four prior decisions and renders the final BUY / SELL / HOLD / CLOSE call with a conviction score.
Configurable per council. Common cadences are 15 minutes, 1 hour, 4 hours, and 1 day. Between full council meetings, a lightweight "check-in" mode runs the Lead Trader on its own at a shorter interval for moment-to-moment HOLD / CLOSE calls within the directional envelope set by the last full meeting.
A full meeting runs all six agents across three tiers. A check-in runs only the Lead Trader, briefed with the standing guidance from the last full meeting plus the latest market data, to catch moment-to-moment changes that don't warrant a full re-deliberation.
Live OHLC streams from Binance International over WebSocket, with automatic Kraken fallback after three failed connections or a region block (HTTP 451). REST backfills cover any gaps. RSI, ADX, +DI, -DI, and ATR are computed on the same timeframe the council is configured for.
A freshness guard fails the council run loudly if the newest OHLC candle for the requested (ticker, interval) is more than three interval-lengths old. Stale data never reaches the agents — the run is marked failed and the next scheduled run picks up once data is flowing again.
Any ticker available on Binance International or Kraken today, with broader coverage (including tokenised equities) on the roadmap. The free tier ships with curated councils; paid users can spin up a council on any ticker and timeframe combination.
Each simulated position carries a take-profit percentage and a trailing-stop percentage. The trailing stop ratchets in the favorable direction only — it never moves against the position. Optional emergency stops close the position regardless of trailing-stop state. Monitoring runs every 60 seconds.
Accuracy depends on the configured agents, models, and market conditions — there is no fixed number. The Council Evaluator runs an LLM auditor over a council's recent calls and proposes prompt edits, so you can iterate. AF7 publishes every simulated outcome in the dashboard, including losing ones.
Yes. Every completed run is stored as an AICouncilExecution row with the final call, conviction score, agent responses, and links to any simulated positions it generated. Run the Evaluator to get an LLM-generated assessment with recommended prompt changes.
Every simulated trade — win, loss, or breakeven — is written to a permanent ledger and attributed to the meeting that opened it, so nothing is cherry-picked. The Council Performance dashboard reports net and realized P&L, win rate, profit factor, maximum drawdown, and trade count, plus an equity curve over time so you can see the shape of the returns rather than a single headline number. It also breaks performance down by conviction — grouping trades into low / medium / high confidence buckets — so you can check whether the council's certainty actually tracks its results. AF7 focuses on P&L, win rate, profit factor and drawdown rather than a single summary ratio, and publishes the losing trades alongside the winners.
Yes — that is exactly what Council Battle does. It clones a council into two head-to-head copies that are identical except one dimension: a 'Battle of Models' keeps the prompts the same and changes the models, while a 'Battle of Prompts' keeps the models the same and changes the prompts. Each side trades its own fresh paper wallet over the same window on the same market, and you watch their equity curves, P&L, win rate, profit factor and drawdown side by side — then replay the whole run. It turns 'which model (or prompt) is better for this strategy' from an argument into a measured result.
Four: AI Council calls (Lead Trader output), candle pattern matches, grid-level crossings, and external TradingView webhooks. All four sources flow through the same processing pipeline, so handling is identical regardless of the source.
Yes. Every model — strategies, councils, positions, signals — is scoped to your user via a tenant system that auto-injects user_id on insert. Other users cannot see your configurations, calls, or simulated positions.
Account top-ups are processed in-app. See the Account page once logged in. Costs are debited per council run, per check-in, and per evaluation, based on the provider's per-token rate plus a small platform margin.
An AI trading council is a group of specialized AI agents that analyze the same market from different perspectives and produce one collective trading signal. In AlphaFlowSeven's implementation, six agents work in three tiers: Technical, Momentum, and Risk analysts read the market in parallel; Bull and Bear advocates then build the strongest case for each direction; finally a Lead Trader weighs the debate and renders a single BUY, SELL, HOLD, or CLOSE call with a conviction score. The structure exists to counter single-model bias — no one model, prompt, or perspective decides alone.
Only if you build them to — and AF7 lets you both engineer and measure that independence. The strongest lever is model diversity: seat different model families (Claude vs GPT vs Gemini vs Grok…) so their failure modes decorrelate, rather than six copies of one model repeating the same mistake. On top of that, the Tier 2 Bull and Bear advocates are explicitly prompted to argue opposite sides before the Lead Trader decides, and AF7 tracks an 'echo chamber' signal — how often both advocates land on the same high-confidence view — so a council that has collapsed into a single voice is visible rather than hidden. And because Council Battle measures whether two model line-ups actually diverge in P&L, diversity becomes testable instead of assumed.
No. AF7 never places orders on any exchange and never connects to an exchange account — there are no exchange API keys anywhere in the product. It is a signals, research, and backtesting platform: the AI council publishes its call and reasoning, a built-in simulation shows how that call would have played out, and any real-world action is entirely yours.
Two ways in AF7. Forward-testing: attach any council to a simulated portfolio and every signal is executed as a paper trade with take-profit and trailing-stop handling, monitored every 60 seconds. Historical backtesting: the grid and candle-pattern engines replay strategies against stored OHLC data using the same crossing-detection logic as live monitoring, so backtest and live behavior match. Both approaches publish full results — including losing trades.
There is no single best model — and AF7 is one of the few platforms that lets you test that question directly. Each of the six council seats can run a different model (Claude, GPT, Gemini, Grok, DeepSeek, Kimi, GLM), and the performance dashboard attributes simulated P&L per council, so you can run identical councils on different models and compare.
The field splits into three categories: social/copy-trading signal feeds, single-model AI analysis tools, and multi-agent research platforms. AlphaFlowSeven is in the third category — its differentiators are the 6-agent debate structure (adversarial bull/bear advocacy before any call), per-seat model choice across seven LLM providers, a transparent simulation that publishes every outcome including losses, and a free tier with three councils and no credit card. Evaluate any platform in this space on: transparency of losing calls, whether reasoning is shown, and whether you can verify performance yourself.
No system — AI or otherwise — reliably predicts prices, and claims to the contrary are a red flag. What multi-agent AI analysis can do is process technical indicators, momentum, and risk factors systematically and consistently, make its reasoning inspectable, and quantify its own conviction. AF7 treats every signal as a hypothesis to be tested in simulation, not a prediction to be trusted, and publishes the full track record either way.
On AF7's free tier: nothing — three councils, full 6-agent pipeline, no card. Beyond that, billing is usage-based: a full 6-agent meeting typically costs a fraction of a cent to a few cents in model fees depending on which LLMs you seat, and lightweight Lead-Trader check-ins between meetings cost less. You pre-fund a balance and pay per run — no subscription.
The fastest way to evaluate AF7 is to sign up and run a council yourself — free for three councils, takes a couple of minutes, no card required.