Splyntra
Observability · Structured Logging

Trace-correlated logs for autonomous agents

Stop hunting across separate log aggregators. Splyntra binds stdout/stderr logs and structured events directly to agent span IDs, making it instant to jump from an agent error to the exact reasoning step that failed.

logger.py
python
import logging
from splyntra.logging import SplyntraLogHandler

logger = logging.getLogger("agent.support")
logger.addHandler(SplyntraLogHandler())

# Injected automatically with active trace_id & span_id
logger.info("Executing retrieval query", extra={"query": "invoice_2026"})
logger.warning("Tool timeout fallback triggered", extra={"latency_ms": 4200})
0ms
Trace Correlation
Logs automatically inherit active span context.
100k/s
Ingestion Throughput
Built for high-volume agent clusters.
Full Text
Log Search
Search across messages, JSON fields, and tags.
Live Stream
Console Tail
Real-time websocket tailing of active agent runs.

Engineered for high-throughput autonomous agents

Every capability is built into the OpenTelemetry streaming pipeline with sub-millisecond ingestion overhead.

01Contextual Logs

Automatic Span Context Injection

Every log line emitted inside an agent function automatically inherits trace_id, span_id, agent_name, and session_id.

  • Seamless integration with Python logging, Winston, Pino, and Logrus
  • 1-click jump from log entry to visual trace timeline
  • Filtered log view for individual sub-agent spans
02Log Querying

Structured JSON Log Explorer

Filter and aggregate logs by custom JSON attributes, user IDs, model parameters, and error codes.

  • High-speed full-text search with regex support
  • Saved query views and alerting on error spikes
  • Time-range correlation across all agents in an organization
03Live Tail

Real-Time Terminal Streaming

Watch live agent execution logs in real time as autonomous multi-agent loops plan, retrieve, and execute tools.

  • Websocket-powered live log stream in the dashboard
  • Colorized log levels (DEBUG, INFO, WARN, ERROR, FATAL)
  • Automatic redacting of sensitive environment variables and tokens

Frequently Asked Questions

How do logs attach to traces automatically?
The Splyntra logging handler uses OpenTelemetry context propagation to extract the active span_id and trace_id from the async execution context (Python contextvars / Node.js AsyncLocalStorage) and injects them into log records.
Can I send standard JSON logs to Splyntra?
Yes. You can send logs via OTLP Logs protocol over HTTP or gRPC, or use our official logging handlers for Python and TypeScript.

Ready to monitor and secure your AI agents?

Get started in under 3 minutes with zero credit card required. Free tier includes up to 5 projects and community telemetry.