Short answer
VaultCharts includes an in-app trading assistant powered by a large language model (LLM) you configure yourself (bring your own key). It is agentic: it can call tools against your charts, loaded OHLC, patterns, pipeline, portfolio, notes, and trade plans. It does not invent live prices from training data, place broker orders, or store your prompts on VaultCharts servers.
VaultCharts itself is free — including the assistant. You only pay your LLM provider if you use a cloud model (or run a free local model via Ollama / LM Studio).
What “agentic” means here
Generic chatbots answer from memory. VaultCharts Assistant can plan multi-step work: fetch or refresh ticker data → scan patterns → compare setups → draft a pipeline card or reflect on journal history. When a step would write something (save memory, update a card, create a planned trade), the app asks you to confirm. Analysis stays in chat; installs and writes stay intentional.
How the assistant works (tools, not vibes)
The assistant is built for analysis and reflection, not financial advice. Typical tool capabilities include:
- Fetch & refresh — pull ticker data from your dataflows / providers.
- Market & OHLC context — samples and market-phase style context for the chart symbol and timeframe.
- Pattern scan — Wyckoff, BOS/CHoCH, order blocks, classic structures, and other supported detectors on the series already in your vault.
- Indicators & strategies — snapshots on loaded series.
- Setup evaluation & comparison — full stock / setup review and ticker comparison workflows.
- Supply & demand / multi-timeframe structure — structure-oriented workflows when relevant.
- Pipeline board — list, rank (e.g. proximity / momentum), draft, and update cards (writes require confirmation).
- Portfolio & risk — holdings review, rebalance suggestions, risk-style reports from your app state.
- Trade plans, notes & journal — open plans, notes, and closed-trade context for reflection.
- On-device curriculum — trading education when you want concepts, not price guesses.
- Learned memory — optional prefs (e.g. max risk %) that you choose to save on device.
- Custom skills & rules — define how the assistant should behave for your process.
- Follow-ups — suggested next questions after meaningful tool runs.
Always treat outputs as educational; verify levels and decisions against your own rules.
Providers and API keys (BYOK)
You choose a provider and model in AI / Models settings (dropdown + model ID, or add a model ID). Supply your own API key for cloud providers. For Ollama / LM Studio, pick the local provider and model from the list.
Supported patterns include:
- Cloud: OpenAI, Anthropic, Gemini, Grok, DeepSeek, Qwen, and similar
- Local: Ollama, LM Studio, or any OpenAI-compatible endpoint on your network
Requests go from the desktop app to the provider you configured. VaultCharts does not sell LLM access, wrap a shared product key, or centrally host inference. Switch profiles in one click and test the connection before you trust it for real sessions.
Example prompts
Try asking things you’d ask a co-trader:
- “Fetch BTCUSDT from Binance, then scan all patterns.”
- “Which pipeline tickers are closest to entry?”
- “Analyze AAPL on the daily and evaluate the setup.”
- “Review my portfolio risk and suggest rebalance ideas.”
- “Remember I prefer max 1% risk per trade.” (asks before saving)
- “What did my last closed trades teach me?”
Chart context and optional @ mentions
You can ask in natural language; the assistant uses the current chart symbol and timeframe when relevant. You can also inject extra context (e.g. market phase or OHLC snippets) via @-style mentions in the composer when you want specific data bundled into a message. See Can I send chart data to the AI?.
Privacy & data flow
- Your API keys are configured locally as yours — not “VaultCharts’ keys.”
- VaultCharts does not store chat prompts or assistant replies for analytics.
- Optional learned memory is saved on device, only after you confirm.
- Market and app data used by tools stay under your control on your machine.
For broader product privacy, see Is VaultCharts private?.
What it will not do
- Provide guaranteed signals, alpha, or outcomes
- Place broker orders or auto-trade your account
- Replace verification of levels against your process
- Invent prices when tools can (and should) read your loaded series
Pricing
VaultCharts — including charts and the AI assistant — is free. There is no Pro unlock for LLM features.
You bring your own API key for cloud providers, or use a local model at no extra cost. VaultCharts does not sell LLM access.
See Is VaultCharts free?.
Troubleshooting tips
- Connection fails — re-check key, provider, and model id; for local models pick the provider from the dropdown and choose/add the model ID; use Test connection in Models settings.
- Thin answers — make sure the relevant ticker / timeframe is loaded, or ask it to fetch/refresh first.
- Unexpected writes — cancel the confirmation dialog; nothing should persist without your yes.
- Privacy-sensitive work — prefer Ollama / LM Studio so inference never leaves your network.