just now
Trading technology provider Spotware Systems, the company behind cTrader (@cTrader), has launched the cTrader Admin MCP server in cTrader version 10.4, allowing brokerage managers to connect AI applications, including OpenAI’s ChatGPT Desktop and Anthropic’s Claude, to their broker data.
Using natural-language requests, managers can investigate trading events, review exposure and account performance, and inspect platform settings. Access stays within each manager’s existing permissions, and every operation is read-only.
The release takes Spotware’s AI work into the broker back office. Spotware introduced cTrader AI Agent Connect in May, giving traders two official MCP servers for linking AI apps to the platform, and followed it in August with cTrader CLI, a desktop command-line app for running trading operations through natural-language commands.
From manual checks to a single conversation
Finding out why an order was rejected or how a disputed trade was executed usually means checking the order history, the linked deals, the account permissions and the symbol settings, then piecing the picture together. The cTrader Admin MCP server hands that job to an AI assistant, which collects the relevant records and prepares an analysis. The manager then reviews the evidence, identifies any further checks and decides what action is needed.
Trading incidents and client complaints
For a client complaint about a rejected order, the assistant retrieves the order history and linked deals, reconstructs the sequence of events and checks it against the account permissions, symbol settings and trading conditions. For a disputed execution price, it compares the recorded deal with the order timeline, pricing settings and historical market data. Where the evidence does not point to an exact cause, the assistant states what remains uncertain and which checks are still needed.
Risk and exposure
Risk reviews work the same way across the whole client base. A manager can ask which symbols currently carry the largest exposure and which accounts contribute to it. The assistant breaks down bought, sold and net volumes by symbol, currency and book, treats A-book and B-book separately, and converts notional values into a single currency for comparison. For open positions, it identifies the largest unrealised gains and losses and traces them back to the accounts behind them.
Account reviews
To find the accounts closest to stop-out, the assistant compares equity, used margin, free margin and margin level against the settings of the relevant group. Realised results can be reviewed in the same way, with a manager selecting a period, comparing accounts and looking at weekly trading activity. All answers are based on retrieved account data. The assistant does not monitor accounts automatically and makes no predictions.
Financial investigations
Balance movements can also be examined. The assistant retrieves the charges recorded against a position and checks them against group and symbol settings, swap rates and charging schedules, marking anything the data does not confirm for further verification. It can also review balance and bonus histories alongside trading records to explain deposits, withdrawals, credit adjustments, fees and other recorded movements.
Configuration review
On the platform side, the assistant reads symbol specifications, sessions, limits and markups, compares them with the broker’s requirements and lists any differences in a checklist for manual review. For pricing, it works through historical bars and ticks and flags gaps or other unusual behaviour for further investigation.
Setup and access control
Managers enable the MCP server in cTrader Admin and connect a supported AI application, such as ChatGPT Desktop, Claude Desktop, Claude Code, Codex CLI or another MCP-compatible app. The server runs locally within cTrader Admin, and access to broker data requires an authenticated manager session. Because access follows that manager’s permissions, the assistant cannot retrieve anything outside their authorised scope.
The assistant gathers data from different parts of the platform and reports its findings in one conversation, but it cannot execute trades or change account settings or configuration. Spotware says findings still need to be verified before any operational decision is made.
“Our goal is to make AI genuinely useful for the people running brokerages every day. A manager investigating a trading complaint should be able to ask a question and get the relevant evidence without spending time collecting records from different sections of the platform. With the cTrader Admin MCP server, we’re making broker data easier to work with while keeping access and operational decisions firmly under the broker’s control.” - Irina Olyaeva, Product Manager for cTrader Admin, Spotware
Other firms have also brought MCP into broker workflows, with trading technology provider Your Bourse @YourBourse letting brokers and prop firms connect their chosen AI assistant to its Trade Server, and multi-asset broker PrimeXBT opening client trading accounts to ChatGPT and Claude.
The LIQUIDITYFINDER app
The LIQUIDITYFINDER app is now available to download on the Apple App Store and Google Play, so members can keep up with industry news, insights and announcements like this one from their phone.
Post, share your views and add content to LIQUIDITYFINDER from your phone: the app is now available on the App Store and Google Play.
