Calls

Call history, transcripts, metrics, problem review and outbound calls. 5 tools.

trillet_list_calls

List call history

List recent calls in the workspace (newest first) with time, duration, status, and full billing fields per call: cost (customer charge in USD, includes any warm-transfer charge), chargedDuration, telephonyCost, warmTransferCharge + warmTransferMinutes (flat-rate human-to-human leg after a warm/whisper transfer, stored as its own line item). Filterable by agent, flow, direction, status, call type, date window, and free text. Use trillet_get_call for one call's transcript and analysis, trillet_get_call_metrics for aggregates.

Flags: read-only structured output

Parameters

Structured output

Returned as structuredContent (fields are optional; the same data is in the text result):

trillet_get_call_metrics

Get call metrics (aggregates)

Aggregate call-history metrics for a filter window without paging individual calls: totalCalls, avgDuration (m:ss string), completionRate. For exact minutes per month use trillet_get_usage include_minutes_breakdown=true (last 3 months) or sum durations from trillet_list_calls; for what was actually charged use trillet_list_transactions.

Flags: read-only structured output

Parameters

Structured output

Returned as structuredContent (fields are optional; the same data is in the text result):

trillet_get_call

Get call transcript and analysis

Fetch one call: transcript, post-call analyzedData, summary, recording URL, and the billing breakdown (cost = customer charge incl. warm-transfer component; warmTransferCharge/Minutes split out; agentSegments for multi-agent calls; transferTiming). include_function_calls=true adds the in-call tool/function invocations (including IVR navigator decisions and DTMF sends, invokedBy 'ivr_navigator'). The primary debugging tool.

Flags: read-only structured output

Parameters

Structured output

Returned as structuredContent (fields are optional; the same data is in the text result):

trillet_review_calls

Review recent calls for problems

Read the recent calls of one flow (flow_id), one agent (agent_id) or the whole workspace and return a compact problem digest: per call the outcome, who ended it and why, transfers and how they went (destination + tool reply; an empty reply on a cold transfer means the call was handed off), and the flagged moments (the caller asked for a person but no transfer happened, transfer_failed, transfer_name_mismatch = the agent asked for a destination that is not an exact transfer friendly_name, silent_after_failed_transfer, the agent kept saying it did not understand, the call was very short or ended in silence, voicemail, errors) with the transcript lines around each. Use it for "what went wrong", "how are calls going", "improve the agent from real calls"; then group the findings, cite call ids, and propose or apply fixes. Use trillet_get_call for a full transcript when a flag needs more context.

Flags: read-only structured output

Parameters

Structured output

Returned as structuredContent (fields are optional; the same data is in the text result):

trillet_start_outbound_call

Start outbound call (REAL CALL)

Place a REAL outbound phone call from an agent to a phone number. This dials a real person and consumes call minutes. Always confirm with the user before calling; requires confirm=true.

Flags: writes destructive open-world billed asks for confirmation

Clients that support MCP elicitation show a confirmation form before this tool runs; declining cancels it with no changes. Other clients rely on their own approval prompt and the confirm argument.

Parameters