Support
Most issues can be resolved using the built-in recovery tools available to your AI agent. Follow the escalation steps below before reaching out.
Troubleshooting steps
This is a cloud-hosted multi-tenant service. You do not have access to server logs, processes, or infrastructure. All interaction happens through MCP tools. Follow this escalation path:
-
Ask your agent to call
check_health. This returns the overall status (HEALTHY / DEGRADED / UNHEALTHY), FIX session state, QUOTE session state, kill switch status, and current risk usage. It's the single best diagnostic starting point.Try asking: "Check the connection health" -
If
fixStateis not ACTIVE, ask your agent to callreconnect_trade_session. This tears down the old connection, opens a fresh one, and reconciles positions from the broker. Common after weekends when markets are closed.Try asking: "Reconnect the trade session" -
If live prices are missing or QUOTE session shows as DISCONNECTED, call
reconnect_quote_session. This restores real-time price data and re-subscribes to all previously subscribed symbols.Try asking: "Reconnect the quote session" -
If you need to change your trading mode, risk limits, or FIX server, use
reset_session. This terminates your current session and triggers re-authentication, opening the configuration form in your browser. Your broker positions and journal entries are not affected.Try asking: "Reset my session so I can reconfigure" -
If the above steps don't resolve the issue, disconnect the MCP connector in your client settings (Claude Desktop: Settings → Integrations; claude.ai: Settings → Connections) and reconnect. This triggers a full fresh OAuth flow.
This is a shared cloud service. You do not have SSH, Docker, or server access. Using reconnect_trade_session and reset_session are the correct recovery tools.
Reporting bugs and requesting features
Use the submit_feedback tool to report issues or suggest improvements. Your AI agent can call it directly.
How to submit feedback
Tell your AI agent what to report:
The tool returns a feedback ID. Save it to check for responses later with get_feedback_status.
Tips for effective reports
- Include health output
Call
check_healthfirst and include the full output in your feedback. This gives the platform team the FIX state, session details, and risk status at the time of the issue. - Describe the exact prompt
Include what you asked the AI agent to do, what happened, and what you expected to happen. Vague reports take longer to investigate.
- Check feedback status
After submitting, use
get_feedback_statuswith the feedback ID to check for updates and responses from the platform team. - Rate limit
Feedback is rate-limited to 10 submissions per session to prevent spam. If you hit the limit, disconnect and reconnect to start a fresh session.
Built-in knowledge base
The server includes a full knowledge base accessible through your AI agent. Ask for any topic:
getting-started Getting Started
Connection setup, OAuth flow, client configuration
order-guide Order Guide
Volume formulas, order types, SL/TP workflow
risk-management Risk Management
Risk guards, kill switch, daily limits
limitations Limitations
Known constraints of the FIX protocol
error-codes Error Codes
Full list of error codes with resolution steps
security Security
Credential handling, encryption, session security
service-model Service Model
Cloud architecture, what you can and cannot access
feedback Feedback
How to report bugs and request features
Full documentation
Detailed guides for every client, trading concept, order type, and configuration option.
Common questions
Check the FAQ before submitting feedback — most common issues (locked out, rejected orders, stale positions) are covered there.
Browse FAQ