Insights MCP
The read-only interface an LLM uses to answer questions about your conversation data.
Ask Thunai
Questions and answers are kept as one thread, so the model remembers what you already asked.
Why did voice CSAT drop on Tuesday?
Voice conversations ending negative rose from 11% to 19% on Tuesday, concentrated in Order issues › Late delivery (64 of 118 negative calls). The turning point in most of them is the courier ETA being restated, around 03:10 into the call.
Answers are drawn only from conversations you are allowed to see, with SafeMind redaction already applied.
Tools exposed
What a connected model can call
Conversations with filters for channel, category, sentiment, agent and date range.
channel, category, sentiment, agent, from, to, limit
One conversation with summary, outcome, quality score and key fields.
conversation_id
Topic-grouped transcript with timestamps and per-line sentiment.
conversation_id, redacted (default true)
Volume, handle time, first contact resolution, quality and Assist acceptance.
metric, group_by, from, to
Sentiment over the conversation with turning points.
conversation_id
Free-text search across transcripts and summaries.
query, channel, from, to
Quick view
LiveConversations today
1,284
Voice 742 · Chat 388 · Email 154
Chat volume, last 7 days
- Avg handle time
- 6m 12s
- First contact resolution
- 81%
- Assist acceptance
- 72%