Use cases
AI agent debugging guides for production teams.
Start with the failure mode, framework, or tracing pattern your team is dealing with. Each guide shows what evidence to capture, how to replay the failed path, and where Opswald fits into the debugging workflow.
AI Agent Debugging
Find the first wrong decision across prompts, context, tools, and side effects.
AI Agent Tracing
Trace production agent runs from prompt to tool side effect.
AI Agent Replay
Reproduce failed agent runs with pinned context and safe tool stubs.
Debug Tool Calling Failures
Inspect schemas, arguments, outputs, retries, permissions, and mutations.
AI Agent Tool Calling
Debug the boundary where model reasoning meets production APIs.
LangChain Agent Debugging
Debug chains, retrievers, memory, callbacks, tools, and retries.
CrewAI Debugging
Trace multi-agent handoffs, tasks, memory, delegation, and tool calls.
OpenAI Agents SDK Tracing
Inspect tools, handoffs, guardrails, model outputs, and side effects.
OpenTelemetry for AI Agents
Use OTel for correlation while preserving agent debugging evidence.
LangSmith Alternative
Evaluate framework-neutral production debugging and replay workflows.
MCP Debugging
Debug failures across MCP servers, tools, permissions, and context.
Developer docs
Go from guide to first trace.
These implementation docs help developers set up Opswald after they understand the debugging workflow.
Quick Start
Trace your first agent run in minutes with the Opswald proxy.
Traces Overview
Understand how Opswald captures spans, tools, and decisions.
Replay Sessions
Step through failed runs span by span with the context that caused them.
API Reference
Use Opswald's API surfaces when you need implementation detail.