Leverate’s September MCP for Traders launch adds a conversational route from an account query to an approved trade instruction. Announced on 16 September and detailed on 21 September, the service lets an AI assistant read account information and prepare orders for customer confirmation. For brokers, the new interface also creates a record-keeping task across the assistant and trading platform.[2][1]
ChatGPT, Claude or another compatible assistant can become the customer’s order-entry surface. The broker chooses which users receive the connector; its trading and risk engine then accepts or rejects submitted orders under the existing rules.[1]
What Leverate says the connector exposes
| Measure | Value | Context |
|---|---|---|
| Connected account | One live or demo account | The trader controls the token and can disconnect the account.Source 1 |
| Available tools | 15 | Ten read account and market data; five place or manage trades.Source 1 |
| Execution model | Two phases | The assistant prepares and restates the order; explicit approval is required before execution.Source 1 |
| Broker control | Off by default | Enabled per brand and targeted by group, trading terms, country or manual allowlist.Source 1 |
| Risk treatment | Existing platform rules | Manual and AI-originated orders pass through the same execution and risk layers.Source 1 |
| CRM record | AI-originated source flag | The broker can identify activity created through the connector.Source 1 |
Where transaction authority sits
Before approval, the assistant is designed to restate the instrument, direction, size, price, stops and margin effect. That creates a specific transaction instruction for the customer to review. Approval records transaction authority; suitability, appropriateness and the quality of the underlying decision remain separate assessments.[1]
Leverate says the assistant explains terms and account data while declining investment advice, recommendations and predictions. In a deployment assessment, that claimed boundary can be examined through representative prompts and languages, alongside the order-confirmation process.[1]
The path from question to market
| Stage | What the AI can do | Control that remains outside the model |
|---|---|---|
| Read | Retrieve balance, margin, positions, orders, prices and instrument details | Scoped token and one-account connectionSource 1 |
| Explain | Translate account and contract information into plain language | Product wording, source data and broker disclosuresSource 1 |
| Prepare | Construct a proposed order and restate its terms | Tool schema, product eligibility and account limitsSource 1 |
| Approve | Present the order for explicit user confirmation | Human decision in the connected AI applicationSource 1 |
| Execute | Send the approved order to the trading platform | Existing execution, risk, challenge and daily-loss controlsSource 1 |
| Record | Flag the order as AI-originated | CRM, account history and dispute trailSource 1 |
The July and September products solve different problems
The July MCP for Brokers product gives authorised employees and operators access to CRM, Broker Portal and trading-platform data through natural-language questions and permitted operational workflows. Firm identity, permissions and data access control that internal service. September’s trader product instead connects one customer account and adds tools to place or manage trades, making customer approval and order evidence central.[3][1]
Deployment scope and the transaction record
Leverate describes MCP for Traders as live on its white-label platform and Client Portal on desktop and mobile, with MT4 and MT5 routes to follow. Brokers enable it by brand and cohort, with jurisdiction-dependent availability. The relevant launch scope is therefore a defined account, platform, country and set of trading terms.[1]
The CRM’s AI-origin flag identifies how an order arrived. To investigate a dispute, a broker would also need to reconstruct the proposed instruction, approval, submitted order, platform response and applicable account rules. The published description leaves open how the assistant-side and platform-side records are joined and retained.[1]
A practical deployment assessment would follow an order through a changed quote, a risk-engine rejection, a repeated request and a revoked connection. These are implementation scenarios for evaluating the documented workflow, with no claim that such tests were performed. They would show how the approved instruction is preserved—or rejected—as it crosses into the broker’s execution system.[1]
Sources & further reading
- MCP for Traders: trade by chat with an AI trading assistant Leverate · 21 Sep 2026 · Accessed 15 Oct 2026
- Leverate launches MCP for Traders to connect AI assistants with trading platforms Leverate / GlobeNewswire · 16 Sep 2026 · Accessed 10 Oct 2026
- Leverate MCP for Brokers connects AI assistants to broker data Leverate · 22 Jul 2026 · Accessed 10 Oct 2026
- Leverate releases MCP for Traders FX News Group · 16 Sep 2026 · Accessed 10 Oct 2026