Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.langsight.dev/llms.txt

Use this file to discover all available pages before exploring further.

​
What is LangSight?

LangSight is the runtime reliability layer for AI agent toolchains. Langfuse watches the brain — prompts, completions, evals, token costs. LangSight watches the hands — the tools your agents call, their health, safety, cost, and blast radius. When your agent fails at 2 AM, LangSight tells you which tool broke, why it broke, how many agents are affected, and stops it from happening again.

​
Where LangSight fits

QuestionBest tool
Did the prompt/model perform well?LangWatch / Langfuse / LangSmith
Is my server CPU/memory healthy?Datadog / New Relic
Which tool call failed in production?LangSight
Is my agent stuck in a loop?LangSight
Is an MCP server unhealthy or drifting?LangSight
Is an MCP server exposed or risky?LangSight
Why did this session cost 47insteadof47 instead of 3?LangSight
If this tool goes down, which agents break?LangSight
Use LangSight alongside Langfuse and LangWatch — not instead of them. They never overlap.

​
The four pillars

​
1. Prevent — stop failures before users notice

from langsight.sdk import LangSightClient

client = LangSightClient(
    url="http://localhost:8000",
    loop_detection=True,        # same tool+args 3x → auto-stop
    max_cost_usd=1.00,          # hard budget limit per session
    max_steps=25,               # hard step limit
    circuit_breaker=True,       # auto-disable tools after 5 failures
)
  • Loop detection — same tool called with same args 3x → session terminated, alert fired
  • Budget guardrails — max cost / max steps per session → hard stop before bill shock
  • Circuit breaker — tool fails 5x → auto-disabled for cooldown → alert → auto-recovery

​
2. Detect — see what broke and why

$ langsight sessions --id sess-f2a9b1

Trace: sess-f2a9b1  (support-agent)  [LOOP_DETECTED]
├── jira-mcp/get_issue        89ms  ✓
├── postgres-mcp/query        42ms  ✓
├──  → billing-agent          handoff
│   ├── crm-mcp/update    120ms  ✓
│   └── slack-mcp/notify    —   ✗  timeout
Root cause: slack-mcp timed out at 14:32 UTC
  • Action traces — every tool call in every session, with latency, status, cost
  • Multi-agent trees — full call tree across agent handoffs
  • Run health tags — every session auto-classified: success, success_with_fallback, loop_detected, budget_exceeded, tool_failure, circuit_breaker_open, timeout, schema_drift

​
3. Monitor — MCP health + security

$ langsight mcp-health

Server              Status    Latency     Schema    Circuit
snowflake-mcp       ✅ UP     142ms       Stable    closed
jira-mcp            ❌ DOWN   —           —         open (5 failures)
postgres-mcp        ✅ UP     31ms        Changed   closed

$ langsight security-scan

CRITICAL  jira-mcp        CVE-2025-6514  Remote code execution
HIGH      postgres-mcp    OWASP-MCP-04   No authentication configured
  • MCP health checks — continuous ping, latency, uptime tracking
  • Schema drift detection — tool schemas change → alert fires before agents hallucinate
  • Security scanning — CVE (OSV), OWASP MCP Top 10, tool poisoning, auth audit

​
4. Map — blast radius via lineage

postgres-mcp ❌ DOWN

Impact:
  - support-agent: 200 sessions/day (HIGH)
  - billing-agent: 50 sessions/day (MEDIUM)

Total: ~250 sessions/day affected
Circuit breaker: active
  • Lineage DAG — which agents call which tools
  • Blast radius — if this tool goes down, what else breaks?
  • Impact alerts — enriched with affected agents and session counts

​
What you can observe per tool type

Tool typeTraceHealth checkSecurity scanCostGuardrails
MCP serversYesYesYesYesYes
HTTP APIsYes——YesYes
Python functionsYes——YesYes
Sub-agentsYes——YesYes
MCP servers get extra depth (health checks, security scanning, schema drift) because the MCP protocol is standard and inspectable.

​
How it works

Your agents                LangSight
──────────                 ──────────────────────────────────────
LangGraph       ──SDK──►   Prevent: loops, budgets, circuit breakers
CrewAI          ──SDK──►   Detect:  traces, health tags, root cause
Pydantic AI     ──SDK──►   Monitor: MCP health, security, schema drift
Claude Desktop  ◄──health─ Map:     lineage, blast radius, impact
                ◄──scan──  Alert:   Slack, OpsGenie, PagerDuty
LangSight dashboard — sessions, error rate, latency overview

​
Installation

pip install langsight
langsight init          # auto-discovers MCP servers
langsight sessions      # view recent sessions with health tags
langsight mcp-health    # check MCP server health + circuit state

Quickstart →

Get from install to your first guarded agent trace in 5 minutes.

​
Integrations

FrameworkIntegration
LangGraphLangSightLangGraphCallback
LangChain / LangflowLangSightLangChainCallback
OpenAI / Anthropic / Gemini SDKwrap_llm()
CrewAILangSightCrewAICallback
OpenAI Agents SDKLangSightOpenAIHooks
Anthropic / Claude Agent SDKAnthropicToolTracer
Pydantic AI@langsight_tool decorator
Claude Desktop / Cursor / VS CodeAuto-discovered by langsight init
Any OTEL frameworkOTLP endpoint

​
Open source

LangSight is Apache 2.0 licensed — free to use, modify, and distribute. 
git clone https://github.com/LangSight/langsight
LangSight session graph — multi-agent call tree showing coordinator, sub-agents, and MCP servers