Get task usage by ID
Get detailed usage data for a specific task by its ID. Returns aggregated metrics across all streams within this task, including token counts, costs, tool calls, and execution times.
Access control:
- Solo users (not in a team): Can view their own tasks only
- Team admins: Can view any team member’s tasks
- Team members/viewers: No access (returns 403)
Authorizations
JWT authentication token
Path Parameters
The unique task identifier
Response
Task usage data
Total agent triggers
When the task was created
LLM model used
Execution status
completed, error, failed, aborted, blocked Number of steps executed
Task ID
Total tool calls made
Billing/model source used by the task. AI connections never consume Neo credits.
neo, byok, ai_connections Agent trigger counts by agent name
Email of the user who created the task (only visible to team admins)
Error message if status is error
Input tokens reported by the upstream provider. Omitted when unavailable.
Deprecated compatibility field. True only when usage_source is byok.
Sum of leaf tool execution time, excluding trigger_* (nested subagent) wait.
Wall-clock union of overlapping LLM call intervals. Never exceeds total_execution_ms.
Provider LLM cost in USD. Present for direct BYOK tasks; omitted for platform-billed tasks and AI connection tasks. Connected subscription providers do not currently report their per-request charge. Preserves up to six decimal places when present.
Sum of pure per-step provider round-trip latencies, excluding tool execution. Additive: overlapping/nested calls do not cancel out, so this can exceed total_execution_ms.
Neo platform cost in USD for direct BYOK usage, llm_cost multiplied by the account's neo_cost_rate. Omitted for platform-billed and AI connection tasks. Preserves up to six decimal places when present.
Neo credits charged for platform-billed usage. Present only when usage_source is neo; omitted for BYOK and AI connection tasks. Task and stream usage preserves up to six decimal places so sub-cent usage remains visible.
Output tokens reported by the upstream provider. Omitted when unavailable.
Average pure provider latency per LLM step.
Median (p50) pure provider latency across this task's LLM steps.
p95 pure provider latency across this task's LLM steps.
Reasoning tokens reported by the upstream provider. Omitted when unavailable.
Number of streams aggregated (1 if single stream)
Stream ID (only present for individual streams, omitted for grouped tasks)
Task title when available
Wall-clock union of overlapping leaf-tool call intervals.
Tool call breakdown with calls and duration
Detailed tool call history
Total execution time (wall clock) in milliseconds
Total reported input and output tokens. Omitted when unavailable.
Sum of time spent waiting on trigger_* (nested subagent) calls. Nests around, not adjacent to, leaf_tool_sum_ms — do not add the two together to derive a wall-time total.

