add_trained_answer
ChatGPTAdd a user-approved question/answer pair to this workspace's SuperBot training. This changes future visitor-facing answers. Confirm the exact text with the user and check search_knowledge first. Creates an entry; never edits or deletes existing training. Repeating the same call does not create a duplicate.
fetch
ChatGPTRetrieve one support conversation by the id returned from search, list_recent_conversations, list_unanswered_questions or list_training_gaps. Returns up to 500 messages and 60,000 transcript characters, with status/sentiment/CSAT metadata, truncation flags and a dashboard URL for the complete history.
get_insights
ChatGPTAnalyst view of this workspace's support over a window of days, compared with the previous window of the same length: top recurring topics customers ask about (with trend up/down/new), repeated questions, conversation and escalation volume change, AI fallback rate (replies where the AI had no grounded answer), CSAT and negative-sentiment change, which sources the AI answered from, and a daily volume series. Use it for 'what are customers asking about', 'what's trending', or a weekly/monthly support review. Topics are keyword themes from visitors' opening messages, not a classifier; read examples with fetch before drawing conclusions.
get_support_overview
ChatGPTAggregate support health for this workspace over a window of days: conversation volume, open/pending/closed, average CSAT with sample size, sentiment split and busiest channels. Use it to answer 'how is support doing'. For topics, trends versus the previous period and AI answer quality, use get_insights.
list_knowledge_sources
ChatGPTList the knowledge sources this workspace's SuperBot agent learns from (websites, documents, files and other ingested content) with type, status, entry count and last sync time, plus a count of trained Q&A answers. Use it to audit what the AI knows, spot failed or stale sources, or answer 'what have we taught the bot'. To search inside the content, use search_knowledge.
list_leads
ChatGPTList sales leads captured in this workspace's SuperBot CRM, newest first: visitors who shared contact details in chat, workflow/form submissions and manually added leads, with name, email, phone, company, source, status, score and notes. Filter by status, source, a text query (name, email or company) or a recent window. Use it for 'who asked for a quote this week' or a lead follow-up review. Contact details are personal data: summarize for the user's team and do not paste them anywhere public.
list_recent_conversations
ChatGPTList the most recent support conversations in this workspace, newest first, optionally filtered by status. Use it for 'what came in today' rather than searching for specific words.
list_training_gaps
ChatGPTFind questions the AI could not answer well, with evidence: visitor questions that got an AI fallback reply (no trained answer, knowledge or model answer was available), conversations escalated to a human, and conversations rated 1-2 CSAT. Returns one item per question with its signals, plus questions that repeat across conversations, plus answer_feedback: how many AI answers visitors marked helpful / not helpful in the widget over the same window. Use it to decide what to teach next: check search_knowledge for an existing answer, then draft one for the user's approval before add_trained_answer. Greetings and very short messages are excluded.
list_unanswered_questions
ChatGPTList recent conversations that may still need an answer, for triage: conversations escalated to a human first, then conversations still open or pending. Each item is a lead to check, not proof the AI failed: read the transcript with fetch and check search_knowledge before proposing training. For questions the AI demonstrably could not answer (fallback replies, escalations, low CSAT) grouped by question, use list_training_gaps.
search
ChatGPTSearch this SuperBot workspace's customer support conversations by words in the subject, visitor name/email or any message. Returns matching conversations (newest first) with an id, title and dashboard URL. Use fetch with a returned id to read the full transcript. For 'what came in recently' without keywords, use list_recent_conversations instead.
search_knowledge
ChatGPTSearch what this workspace has taught its SuperBot agent: trained question/answer pairs and ingested knowledge sources. Use it to check whether an answer already exists before writing a new one. To list all knowledge sources and their sync status, use list_knowledge_sources.
add_trained_answer
Claudefetch
Claudeget_support_overview
Claudelist_recent_conversations
Claudelist_unanswered_questions
Claudesearch
Claudesearch_knowledge
Claude