StackMap
Subscribe

agentsight vs LangSmith

System-level observability for AI agents via eBPF and TLS tracing: correlates prompts and model calls with the processes, files and network the agent actually touched — no SDK, no proxy. — versus — Trace, test and monitor LLM apps in production.

The curated verdict

AgentSight's own framing: LangSmith is right when you own the application code and want traces, prompts and evals; AgentSight is for the layer below — a closed agent CLI whose real effects (processes, files, network) only show up at the system boundary.

agentsightLangSmith
Stars6710
Forks101
LanguageCTypeScript
LicenseMITProprietary
Last activitytoday1 week ago
Topicssecurity, coding, localevals
Curated connections55

agentsight — the curator's take

Use it when the agent is a closed CLI (Claude Code, Codex, Gemini CLI, OpenClaw) and its own logs don't explain a failure: AgentSight watches at the kernel boundary, so it sees the model traffic *and* every exec, file write and outbound connection, and ties them together. That's the layer LangSmith-class tools can't reach without owning the app code. NOT for prompt/eval-level tracing of an app you wrote (use an OTLP tracer), and it's Linux-with-eBPF: on macOS you're out of luck. Not an enforcement tool — it observes and advises, it doesn't block.

LangSmith — the curator's take

Trace, test and monitor LLM apps in production.