Ai Trader MT5 — Hilfe
VibeFox AI Trader is an Expert Advisor for MetaTrader 5 that uses artificial intelligence to analyze charts and generate trading signals. Send a chart screenshot to the neural network — and receive a detailed analysis with trend lines, support/resistance levels, and trading signals directly on the chart.
HOW TO CHOOSE AN AI MODEL
The Expert Advisor will not work without an external model, so you need to choose the right model.
The model must be able to process images — VL.
It is highly recommended that the model has at least 8B parameters for analysis or from 30B for analysis and trading. The bigger — the smarter. The ideal option is paid models or models from 100B.
At the moment, it is not recommended to trade on a real account, because there are not yet enough capable AIs. It is recommended to use it for analysis and advice.
Local AI Models
Local models are free. We recommend working specifically with them.
How to choose a local model depending on hardware
If you have a very weak video card, it is possible to use only RAM, but the speed will be slow. Dense models larger than 8B are practically impossible to use. For large models, choose only MOE. The model must have at least q4 quantization, and the more, the better. Video cards with a large amount of memory will significantly speed up the work. A Dense model works well only when it fully fits into video memory along with the KV cache.
Qwen VL models show the best results — verbose analysis and order placement. Qwen3 VL 30B a3b is the best of the available models for a medium computer.
RAM only:
| 8 GB RAM | 4B models. Very weak analysis, often unable to create orders. |
| 16 GB RAM | 8B models. Image processing will take more than 1 minute. Analysis quality is weak. |
| 32 GB RAM | 30B MOE models. Qwen3 VL 30B MOE — a very good option. Processing may take several minutes. |
| 64 GB RAM | 30B–70B MOE models. Processing may take several minutes. |
Video card and RAM:
| 8 GB VRAM + 16 GB RAM | 8B models will process an image in about 20–30 seconds. |
| 16 GB VRAM + 32 GB RAM | 30B moe models will process an image in about 20–60 seconds. |
| 16 GB VRAM + 64 GB RAM | 70B moe models will process an image in more than 1–2 minutes. |
Examples of models compatible with LM Studio/Jan.ai:
<7B Not recommended. Very fast, but weak analysis. Almost unable to place orders.
| Qwen3 VL 4B | the best of the 4B for quick analysis, but not for orders. |
| Gemma 3 4B | the fastest |
8-14B Fast. For simple analysis. They place orders, but should not be trusted.
| Qwen3 VL 8B | |
| Minicpm v4.5 8B | |
| Glm 4.6V flash 9B | |
15-40B Recommended for 8 GB VRAM + 32 GB RAM. Good chart analysis and order placement.
| Qwen3 VL 30B a3b moe | the best model for local launch on a mid-budget machine. |
40-110B At the moment, we have not found good VL models.
| GLM 4.7V 70B moe | long and powerful analysis, but poor at drawing on the chart and placing orders. Use only for analysis. |
>110B Recommended for unlimited capabilities
Premium AI Models
For paid models, you need to find out the price for processing one image experimentally. Prices often change and vary greatly between different companies, models, and subscriptions. For example, processing with ChatGPT's latest models costs more than $1 per image. The exact price can only be found out by making 1 request and seeing how much money was spent on it. Current approximate prices can be found in AI chats, but they are not always accurate. We recommend using Chinese, cheaper models — GLM, DeepSeek, MiniMax, Mimo.
INSTALLATION
Required MT5 settings (Tools -> Options):
Expert Advisors tab:
- [x] "Algo Trading" in the options or on the toolbar — allow trading. If you will only use analysis, it is not required.
- [x] "Allow WebRequest for listed URL" — allow web requests. Required for communication with AI. Add the URLs of your AI providers and localhost for local models:
| Local | http://127.0.0.1 (for all local models) |
| Claude (Anthropic) | https://api.anthropic.com |
| Gemini (Google) | https://generativelanguage.googleapis.com |
| OpenAI | https://api.openai.com |
| OpenRouter | https://openrouter.ai |
| Together AI | https://api.together.xyz |
| Fireworks AI | https://api.fireworks.ai |
| DeepSeek | https://api.deepseek.com |
| Mistral AI | https://api.mistral.ai |
| Groq | https://api.groq.com |
| Perplexity | https://api.perplexity.ai |
| Kimi / Moonshot | https://api.moonshot.ai |
| MiniMax | https://api.minimax.io |
| GLM / Zhipu AI | https://api.z.ai |
Optional:
- Notifications tab — Enable Push Notification — for sending Push messages
- Email tab — Enable Email Notification — for sending Email messages
- FTP tab — enter FTP credentials — for saving chart analysis to FTP
PANEL AND CONTROLS
Panel
The panel is divided into sections. In the settings, all sections can be enabled/disabled and made open/closed by default. Please note that if a section is closed or disabled, the options on it continue to work. You can quickly close and open all sections by clicking on the icons in the header. You can also configure the panel size and font size. There are several preset modes and your own settings (Custom)
Panel header
- Name and version of the Expert Advisor
- Down arrow icon — Opens all sections
- Up arrow icon — Closes all sections
- Underscore/Square icon — Close/Open the panel
- Cross icon — Close the Expert Advisor
- Double click on the header — Close/Open the panel
- Press and hold to move the panel
Start and stop work
- Start work — Click the [Start] button to begin processing symbols using AI
- Timeout — AI processing time. During the timeout, the panel stops responding to any clicks. This is a feature of MQL5, not the Expert Advisor. Nothing can be done until the result is received. In the settings, you can set the timeout time (default 60 seconds). If the model does not have time to process the result, an interruption will occur. Usually 60 seconds is enough for paid and fast models, but sometimes it may not be enough, for example, for large local models. You can set a longer time. The average response from major AIs (Claude, Google, OpenRouter) is 10–15 seconds.
- Red STOP button — processing is in progress.
- The button displays the current symbol number and the total number of symbols. For example 1/10.
- How to stop the process? Click the red button once, and the process will stop after the timeout! Immediate stop will not happen. As a last resort — close MetaTrader.
- If the scheduler is enabled, the time until launch will be displayed on the Stop button.
Requests to AI (Prompts)
It is much more convenient to keep requests in files than in Expert Advisor settings. The settings can hold up to 5 requests. To work with files, the default text editor for .txt files is used — usually Notepad. All request files are edited in the Metatrader/MQL5/Files/Ai_Trader/Prompts folder. You can create any request files there. Then click Refresh to add the files to the request list.
| Prompt selection | Drop-down list with prompt selection. In the list, file prompts are marked with the symbol [F] at the beginning. |
| [Result] | Shows the link to the result.html file. Copy and paste it into your browser. Save it in favorites and then simply open it — the address is always constant. |
| [Repeat] | Repeat the last response without sending it to AI. All the same actions happen instantly — drawing analysis and orders, sending notifications. |
| [Show] | Show/hide the text of the selected prompt on the chart |
| [Refresh] | Refresh the list. Needed after manually changing prompt files. |
| [Add] | Create a new prompt file. Enter the name without the .txt extension and click OK. |
| [Rename] | Rename a prompt file (file-based prompts only) |
| [Del] | Delete a prompt file (file-based prompts only) |
SYMBOLS
Symbols are iterated one by one from the specified lists. They can be specified in the settings, files, or by selecting collections. Note that if there are many symbols and they are loading for the first time, the cache may not always have time to load. A small timeout is needed for this. It can be set in the settings. All files with symbol lists are edited in the Metatrader/MQL5/Files/Ai_Trader/Symbols folder. You can create any files there. Then click Refresh to add the files to the symbol list.
| Symbol list selection | Symbol list selection (Current — current symbol only, MarketWatch — all open symbols in MarketWatch, Manual — symbols entered in the Expert Advisor settings, files — your own lists from text files) |
| [Show] | Show/hide the symbol list on the chart. |
| [Refresh] | Refresh the symbol list from files. |
| [Add] | Create a new symbol list file. Enter the name without the .txt extension and click OK. |
| [Rename] | Rename a symbol list file (files only) |
| [Del] | Delete a symbol list file (files only) |
RESPONSE
You can choose what to draw on the chart after the AI response.
| [x] Show Analyze | AI draws lines/zones/labels on the chart |
| [x] Show Orders | AI generates trading signals (Entry, Stoploss, Takeprofit lines) |
TRADING
Trading settings.
| Lot: [0.01] | Lot size. Can be changed with the mouse wheel or Up/Down keys. Ctrl/Shift modifiers are taken into account — different digits in the number will change. |
| Real Orders: [0] | Real orders. How many real orders to open (0=off, N=first N symbols). Orders are opened for the first symbols in the list. If 5 is specified, orders will be opened on the first 5 symbols. Orders are not opened if an order is already open on the symbol. |
| [x] Order Edit | Enable manual correction of the entry point, Stop Loss, Take Profit. This works only on the current chart where the Expert Advisor is set. After symbol processing is complete, sliders and buttons appear at the top right. The sliders can be moved up, down, left, and right. The OK button creates an order, the X button cancels it. |
AI MODELS
| AI provider selection | Select the desired provider from the list (Local, Claude, Gemini, OpenAI, etc.). When selected, models are collected and the model list to the right is populated. |
| Model selection | The model list is loaded automatically from the provider. In the settings, you can filter models by release date and keywords. This is needed because without this, too many models are returned that cannot be used. There are many text models that cannot process images. |
NOTIFICATIONS
| [x] HTML | Save results to result.html (text + screenshot). Can be opened in any browser. |
| [x] Telegram | Send to a Telegram bot (text + photo) |
| [x] FTP | Save to FTP (FTP access must be configured in MetaTrader options) |
| [x] Chart (text) | Show AI response as text on the chart |
| [x] Response image only | Send only the screenshot to Results.html/Telegram/Webhook/FTP (without text). For this, you need to enable text output to Chart. Then no additional text will arrive and it will take up less space and traffic. The text will be on the screenshot. |
| [x] Email (text) | Send email notification (text only). You need to enable and configure Email access in MetaTrader options. |
| [x] Push (start) | Push notification at work start. You need to enable and configure Push access in MetaTrader options. |
| [x] Alert (start) | MT5 Alert at work start. |
| [x] Webhook | POST JSON to webhook URL (text + photo base64). Message format: |
{
"message": "EURUSD [gpt-4o]: BUY signal detected",
"image_base64": "/9j/4AAQSkZJRg...",
"request": { ...JSON ... },
"response": { ...JSON ... },
"time": "2026.02.21 14:30:05"
}
VISUAL
Panel appearance settings.
| [S][M][L][XL][C] | Change panel size (300/400/500/600/Custom px). Custom — your own settings via Expert Advisor options. |
| [Dark/Light] | Switch color scheme Light/Dark |
| [Slider S][M][L] | Change slider size in OrderEdit (small/medium/large) |
SCHEDULER
The scheduler is needed to start work at a specific time. After enabling it, click START, and you will see a timer until the nearest work start. There are detailed interval and time settings. On the panel, you can only enable and disable it.
There are 3 types of scheduler:
- Interval — for example, every 5 minutes.
- Interval from the end of work — after the end of work, the interval is measured and the process starts
- Specific time — you can set a specific time for each day of the week.
| [x] Enable | Scheduler enable/disable. |
ACTIONS
Additional actions.
| [Folder] | Show the address of the Ai_Trader data folder. Copy it into Explorer and it will open. |
| [Log] | Show the address of the log.txt file. Open in Notepad. |
| [Help] | Show the address of the online help page. Copy into a browser. |
| [Bug report] | Show the address of the online contacts and bug reports page. Copy into a browser. |
CLEAR
Clearing files, logs, results.html
| [Results] | Clear result.html |
| [Log] | Clear log.txt |
| [Screenshots] | Delete all screenshots from the Screenshots folder |
| [Requests] | Delete all requests from the Requests folder |
| [Objects] | Delete all drawn objects from the chart |
| [Chart] | Clear text on the chart |
| [x] Auto Clear on Start | Clear everything at each work start. Convenient to prevent accumulation. |
DEBUG
Debugging. Tools for analyzing requests and responses.
| [x] Debug | Debug mode. When enabled, JSON Request and Response are added to any analysis text for detailed AI workflow analysis |
| [Last Request] | Open the last API request file in a text editor. |
STATUS
| Status bar | Shows the current state (Ready, processing progress, errors) and a brief description when hovering over panel elements. |
INPUT PARAMETERS
PANEL
| Instance Name | Expert Advisor instance name (empty = auto-generation). Needed if you use more than 1 Expert Advisor in one program. Each one must be given a different Instance Name. |
| Panel Size | Panel size: S(300), M(400), L(500), XL(600), Custom |
| Custom Width | Width in pixels for Panel Size = Custom |
| Custom Font Size | Font size for Panel Size = Custom |
| Color Scheme | Theme: Light or Dark |
SECTIONS
All sections can be made open, closed, or disabled by default. Note that if a section is disabled, the options on it do not work.
| AI Models | Section state: Open / Closed / Disabled |
| Symbols | Section state: Open / Closed / Disabled |
| Process | Section state: Open / Closed / Disabled |
| Trading | Section state: Open / Closed / Disabled |
| Notifications | Section state: Open / Closed / Disabled |
| Visual | Section state: Open / Closed / Disabled |
| Scheduler | Section state: Open / Closed / Disabled |
| Actions | Section state: Open / Closed / Disabled |
| Clear | Section state: Open / Closed / Disabled |
| Debug | Section state: Open / Closed / Disabled |
SYMBOLS
| Symbol Mode | Symbol mode: Current / MarketWatch / All / Currencies / Manual |
| Manual Symbols | Comma-separated symbol list for Manual mode |
| Show "All" option | Show "All" in the drop-down list. If enabled, a very long iteration of a large number of symbols may begin. |
PROMPTS
| Prompt 1-5 Name | Built-in prompt name (empty = skip) |
| Prompt 1-5 Text | Prompt instruction text |
SCHEDULER
| Scheduler Mode | Mode: Time (setting an exact start time) / Interval / Interval from end (interval from the end of work) |
| Interval (minutes) | Timer interval in minutes (minimum 1) |
| Monday-Sunday | Schedule by day: format "HH:MM,HH:MM", for example "10:00,12:00,14:00" |
TRADING
| Lot Size | Default lot size |
| Slippage (points) | Maximum slippage |
OrderEdit
| Slider Size | Slider size for OrderEdit: 0=S, 1=M, 2=L |
| Entry Line Color | Entry line color |
| Entry Line Width | Entry line width (1-5) |
| SL Line Color | Stop Loss line color |
| SL Line Width | Stop Loss line width (1-5) |
| TP Line Color | Take Profit line color |
| TP Line Width | Take Profit line width (1-5) |
AI SETTINGS
| AI Response Timeout | Max AI response wait time in seconds — timeout (30–600) |
Local AI (up to 3 servers)
| Local 1-3 Name | Display name (e.g., "LM Studio", "Ollama") |
| Local 1-3 IP:Port | Server address (e.g., "127.0.0.1:1234") |
Claude
| Endpoint | API URL (default: https://api.anthropic.com) |
| API Key | Anthropic API key (sk-ant-...) |
| Min Date | Exclude models older than YYYY.MM |
| Models filter | Exclude models containing these words (comma-separated) |
Gemini
| Endpoint | API URL (default: https://generativelanguage.googleapis.com) |
| API Key | Google API key (AIza...) |
| Min Date | Exclude models older than YYYY.MM |
| Models filter | Exclude models containing these words (comma-separated) |
OpenAI Compatible 1-4
| Provider Name | Display name in the API drop-down list |
| Endpoint | API URL (must support /v1/models and /v1/chat/completions) |
| API Key | Provider API key |
| Min Date | Exclude models older than YYYY.MM |
| Models filter | Exclude models containing these words (comma-separated) |
NOTIFICATIONS
| Chart (text only) | Show AI response as text on the chart |
| Chart Font Size | Font size on the chart (8-20) |
| Text on Dark BG | Text color on a dark chart background |
| Text on Light BG | Text color on a light chart background |
| Chart Font Bold | Bold font for text on the chart |
| Chart Line Width | Text wrap width in characters (30-60) |
| Chart Text X | Horizontal text offset in pixels |
| Chart Text Y | Vertical text offset in pixels |
| Screenshot Only | Send only the screenshot. You need to enable the Chart (text only) option, then the text will be on the screenshot. Convenient if there is not much text. |
| Alert (start) | Show MT5 Alert at work start |
| Push (start) | Send Push notification at work start |
| Results.html | Save results to an HTML report file |
Telegram
| Enabled | Enable Telegram notifications |
| Bot Token | Telegram bot token from @BotFather |
| Chat ID | Target chat/channel ID |
Email
| Enabled | Enable email notifications |
Webhook
| Enabled | Enable webhook notifications |
| URL | Webhook endpoint URL (receives POST with JSON) |
CLEAR
| Auto Clear on Start | Auto-clear all objects and text before each launch |
DEBUG
| Debug Mode | Extended logging. Full JSON Request/Response is added to the text |
ORDER EDIT
- Order editor after the AI response
- The editor can be enabled and disabled in the settings and on the panel
- 3 sliders appear on the chart — Entry point, Stoploss, Takeprofit.
- They can be moved with the mouse up, down, left, right.
- On the Entry point slider there is an OK button — to enter the position and X to cancel
- There are additional buttons at the top right
| [Reset] | Reset Entry/SL/TP to original values |
| [Text] | Show/hide AI response text on the chart |
| [Analyze] | Show/hide AI analysis graphics on the chart |
| [Order] | Show/hide order lines/sliders on the chart |
FOLDER STRUCTURE
The Expert Advisor creates its own folder structure in MQL5/Files/Ai_Trader. You can access it by clicking Actions->Folder on the panel.
For speed, you can create Prompts and Symbol Lists directly in files, rather than through the panel, for example. The Expert Advisor must be restarted so that it picks up the new lists from the files.
| Prompts | Text files of prompts |
| Requests | JSON files of requests and responses, for analyzing work with AI. |
| Screenshots | Chart images |
| Symbols | Text files with symbol lists. |
| last_message.txt | last message |
| last_request.txt | last request in JSON format |
| last_response.txt | last response in JSON format |
| log.txt | log |
| result.html | file with results for viewing in a browser. |