# LangSight ## Docs - [Agents](https://docs.langsight.dev/agents.md): Browse every agent LangSight has observed — health scores, session counts, topology view, and per-agent reliability metrics - [Prevention Config (Guards)](https://docs.langsight.dev/agents/prevention-config.md): Configure per-agent safety controls — loop detection, budget caps, and circuit breakers — that run inside the LangSight SDK before each tool call. - [Alerts & Notifications](https://docs.langsight.dev/alerts.md): Configure Slack alerts, manage the alert inbox, and understand what fires and when - [Anomaly Detection](https://docs.langsight.dev/anomaly-detection.md): Statistically detect unusual tool behaviour using z-score comparison against a 7-day baseline - [Health API](https://docs.langsight.dev/api-reference/health.md): Query MCP server health status and history - [API Overview](https://docs.langsight.dev/api-reference/overview.md): LangSight REST API — all endpoints with authentication and rate limits - [Security API](https://docs.langsight.dev/api-reference/security.md): Trigger security scans and retrieve findings - [Traces API](https://docs.langsight.dev/api-reference/traces.md): Ingest tool call spans from the SDK or OTLP - [langsight costs](https://docs.langsight.dev/cli/costs.md): Show cost breakdown by LLM model, tool call, and project - [langsight init](https://docs.langsight.dev/cli/init.md): Auto-discover MCP servers from 10+ IDE clients and generate .langsight.yaml - [langsight investigate](https://docs.langsight.dev/cli/investigate.md): AI-assisted failure investigation for MCP servers - [langsight mcp-health](https://docs.langsight.dev/cli/mcp-health.md): Check the health of all configured MCP servers - [langsight monitor](https://docs.langsight.dev/cli/monitor.md): Continuously monitor MCP servers and alert on state changes - [langsight scan](https://docs.langsight.dev/cli/scan.md): Zero-Docker MCP health + security audit in a single command. No Postgres, ClickHouse, or Docker required. - [langsight security-scan](https://docs.langsight.dev/cli/security-scan.md): Scan MCP servers for CVEs, OWASP MCP Top 10 issues, and tool poisoning - [langsight serve](https://docs.langsight.dev/cli/serve.md): Start the LangSight REST API server - [langsight sessions](https://docs.langsight.dev/cli/sessions.md): List agent sessions and display multi-agent call trees - [Model Pricing](https://docs.langsight.dev/costs-model-pricing.md): Token-based LLM cost tracking — pre-seeded models, custom pricing, per-project filtering - [Dashboard](https://docs.langsight.dev/dashboard.md): Overview tab — error breakdown by category, Tools tab — calls per session, tool retry rate, and silent failures, Models tab — context window pressure - [Database Connections](https://docs.langsight.dev/database-connections.md): Connect to PostgreSQL and ClickHouse with DBeaver, psql, or clickhouse-client - [Auto-Discovery](https://docs.langsight.dev/guides/auto-discovery.md): Zero-config agent and server registration from trace data - [Installation](https://docs.langsight.dev/installation.md): Install LangSight on macOS, Linux, or Windows. - [Introduction](https://docs.langsight.dev/introduction.md): Agent runtime reliability — prevent loops, enforce budgets, break failing tools, and map blast radius across your agent toolchains. - [Live View](https://docs.langsight.dev/live-view.md): Real-time monitoring of agent sessions as they run — watch spans tick up, catch errors immediately, and spot hung agents before they waste budget - [MCP Servers](https://docs.langsight.dev/mcp-servers.md): Browse every MCP server your agents call — declared tools, health history, reliability metrics, and editable catalog metadata - [MCP Server Authentication](https://docs.langsight.dev/mcp/authentication.md): Configure authentication for every MCP server transport — from local stdio to OAuth-backed remote servers. - [Blast Radius](https://docs.langsight.dev/mcp/blast-radius.md): When an MCP server goes down, instantly see which agents and sessions are affected — based on real tool-call traffic from the last 24 hours. - [Health Monitoring](https://docs.langsight.dev/mcp/health-monitoring.md): Continuous uptime checks, latency tracking, and status classification for every MCP server in your stack. - [MCP Servers Dashboard](https://docs.langsight.dev/mcp/mcp-servers-dashboard.md): The unified /servers page — health status, tool catalog, latency trends, and agent consumers for every MCP server. - [Project Scoping](https://docs.langsight.dev/mcp/project-scoping.md): How project-scoped API keys automatically route all CLI output to the right project — no flags needed. - [MCP Quickstart](https://docs.langsight.dev/mcp/quickstart.md): Get MCP server health monitoring running in under 5 minutes — from install to live health checks and A-F scorecards. - [Schema Drift Detection](https://docs.langsight.dev/mcp/schema-drift.md): Detect exactly what changed in an MCP server's tool schema — breaking vs compatible — and find which agents are calling the affected tools. - [Server Scorecard](https://docs.langsight.dev/mcp/scorecard.md): A-F composite reliability grade for every MCP server — weighted across availability, security, reliability, schema stability, and performance. - [Server Discovery](https://docs.langsight.dev/mcp/server-discovery.md): How LangSight finds your MCP servers automatically, and how to register servers manually with `langsight add`. - [Tool Schema Viewer](https://docs.langsight.dev/mcp/tool-schemas.md): Browse the full JSON input schema of every tool on every MCP server — parameter names, types, required flags, descriptions, and enum values — directly in the dashboard. - [Projects](https://docs.langsight.dev/projects.md): Project isolation and RBAC — scope all reliability data per project - [Claude (Anthropic)](https://docs.langsight.dev/providers/claude.md): Use Claude for root cause analysis in langsight investigate - [AI Providers](https://docs.langsight.dev/providers/overview.md): Configure Claude, OpenAI, Gemini, or Ollama for langsight investigate - [Quickstart](https://docs.langsight.dev/quickstart.md): Zero to your first agent trace in 5 minutes. - [Anthropic SDK / Claude Agent SDK](https://docs.langsight.dev/sdk/integrations/anthropic.md): Trace tool_use calls from the Anthropic SDK and Claude Agent SDK. - [Anthropic SDK](https://docs.langsight.dev/sdk/integrations/anthropic-sdk.md): Auto-trace Anthropic tool_use blocks with wrap_llm() — one line, zero code changes - [CrewAI](https://docs.langsight.dev/sdk/integrations/crewai.md): Full observability for CrewAI agents — crew, task, agent, tool, LLM, and handoff spans with zero code changes - [LangChain + Langflow + LangGraph](https://docs.langsight.dev/sdk/integrations/langchain.md): Trace every tool call in LangChain agents, LangGraph workflows, and Langflow — one callback, all frameworks - [LangGraph](https://docs.langsight.dev/sdk/integrations/langgraph.md): Full observability and runtime safety for LangGraph agents — topology capture, loop detection, budget enforcement, and thinking token tracking. - [OpenTelemetry (OTLP)](https://docs.langsight.dev/sdk/integrations/otel.md): Point any OTEL-instrumented framework at LangSight — zero code changes - [Python SDK](https://docs.langsight.dev/sdk/python.md): Add 2-line agent runtime reliability to any Python agent - [Backup & Restore](https://docs.langsight.dev/self-hosting/backup-restore.md): Backup and restore procedures for LangSight's PostgreSQL and ClickHouse databases - [Configuration Reference](https://docs.langsight.dev/self-hosting/configuration.md): Complete .langsight.yaml schema - [Docker Compose](https://docs.langsight.dev/self-hosting/docker-compose.md): Production deployment with ClickHouse, PostgreSQL, and OTEL Collector - [Feature Matrix](https://docs.langsight.dev/self-hosting/feature-matrix.md): Which features require PostgreSQL, ClickHouse, or both — at a glance - [Production Deployment](https://docs.langsight.dev/self-hosting/production.md): Deploy LangSight in production — one command starts everything in v0.9.0. - [Production Hardening](https://docs.langsight.dev/self-hosting/production-hardening.md): Nginx reverse proxy, TLS, firewall rules, and security configuration for production LangSight deployments - [Redis](https://docs.langsight.dev/self-hosting/redis.md): Redis setup, architecture, high availability, monitoring, and troubleshooting for LangSight - [Scaling Guide](https://docs.langsight.dev/self-hosting/scaling.md): Current performance benchmarks, tunable settings, and a clear upgrade path from single-node pilot to production-scale deployment. - [Storage Backends](https://docs.langsight.dev/self-hosting/storage.md): Choose between PostgreSQL, ClickHouse, and the dual-backend production topology - [Troubleshooting](https://docs.langsight.dev/self-hosting/troubleshooting.md): Common LangSight startup and runtime errors and how to fix them - [Upgrading](https://docs.langsight.dev/self-hosting/upgrading.md): How to upgrade LangSight between versions without data loss — standard steps, version-specific notes, and rollback procedure - [Session Health & Status](https://docs.langsight.dev/session-health.md): Understanding health tags and status indicators on the Sessions page — what each tag means and how they are computed - [Agent SLOs](https://docs.langsight.dev/slos.md): Define success_rate and latency_p99 SLOs per agent — track whether your agents meet their targets - [User Management](https://docs.langsight.dev/users.md): Invite teammates, assign roles, and manage access ## OpenAPI Specs - [openapi](https://docs.langsight.dev/api-reference/openapi.json) ## Optional - [GitHub](https://github.com/LangSight/langsight) - [Changelog](https://github.com/LangSight/langsight/blob/main/CHANGELOG.md)