Roadmap
What we’re building next. No dates — just direction.
Recently shipped
LLM-as-judge quality scoring for agent outputs; correctness, groundedness, JSON validity, and tone.
Visual pipeline graph in the trace modal with per-agent cost, duration, and status.
Parent agent runs automatically link to child agents without manual parent_trace_id passing.
Detects redundant tool calls, model mismatch, and cost spikes with estimated monthly savings.
Point your OTLP/HTTP exporter at Farol — no SDK changes required.
Cost, latency, and success rate drift detection.
Multiple agents affected by the same issue shown as one alert.
Connect child agent runs to their parent pipeline.
Compare cost and quality across prompt versions.
One number (0–100) combining all agent signals.
LangChain, CrewAI, and Vercel AI SDK examples in the docs.
Coming next
Not just time proximity.
Beyond documentation examples.