search-logs

Search Logs

Use Shipbook MCP tools to search for specific log entries.

Steps

1. Call get-account-apps to identify the target app if not already known (returns name, appId, platform, key, integration). If the user didn't specify an app, detect the current project's platform and filter matching apps. If there are still multiple matches, ask the user which one.
2. Translate the user's natural language query into appropriate get-logs filters:
   • Severity → severity (Error, Warning, Info, Debug, Verbose). When investigating issues, start with Error.
   • Text search → msg (free text search in log messages)
   • Code location → fileName, tag, className, lineNumber
   • Device → udid (unique device ID — use for specific device searches), deviceModel, manufacturer, osName (e.g. iOS, Android), osVersion, language
   • User → userId, userName, fullName, email, phoneNumber
   • App → appVersion
   • Negation → prefix any filter value with ! or - to exclude (e.g. appVersion: !1.5.5)
   • Version comparison → appVersion and osVersion support >, >=, <, <= (e.g. appVersion: >2.0)
   • Loglytics → key (error key from get-loglytics-errors — finds all logs for a specific classified error)
   • Session → session (session ID — retrieves all logs from that session. After finding an error, use its session ID without severity filter to see full context)
   • Pagination → minOrderId (logs after this order), maxOrderId (logs before this order) — use to navigate within a session
   • Time range → fromTime, toTime (ISO format. If user mentions a specific time, set 10-minute window around it)
   • Limit → limit (default: 50)
3. When analyzing a specific log entry, get its surrounding context: call get-logs again with the same session ID, set maxOrderId to the log's orderId + 10, and remove severity filter. This shows what happened leading up to that log.
4. Every log includes fileName and lineNumber — use these to locate and read the actual code in the project. This helps understand what the log means in context.
5. Present results as a readable summary, not raw JSON.
6. If too many results, suggest narrowing filters. If no results, suggest broadening.