LangWatch Claude Code Marketplace
langwatch/claude-marketplace
Agentes especialistas para testes Scenario e observabilidade LangWatch
Ver no GitHubMarketplace de plugins Claude com agentes especializados para testes de agentes Scenario e observabilidade LangWatch. Fornece dois agentes delegados (scenario-expert e langwatch-expert) apoiados pelo MCP server LangWatch, com acesso a documentação ao vivo e traces de execução.
Instalação
npx convez-ai langwatch/claude-marketplace1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Debugging e desenvolvimento de testes de agentes com Scenario
- Análise de traces e analytics de execução em tempo real
- Instrumentação e rastreamento de pipelines de IA com LangWatch
- Investigação de comportamentos anômalos em sequências de mensagens
Veredito

Ferramenta especializada para equipes usando LangWatch e Scenario, integrando observabilidade e testes diretamente no Claude. Sem configuração obrigatória (docs funcionam offline), com suporte a automação via CI/CD.
README
A tradução para o português deste README ainda está na fila.
LangWatch Claude Code marketplace
A Claude Code plugin marketplace for the LangWatch org. One plugin today: langwatch-experts.
Install
/plugin marketplace add langwatch/claude-marketplace
/plugin install langwatch-experts@langwatch
What you get
Two delegate-to expert agents, backed by the LangWatch MCP server:
scenario-expert— writing, debugging, and reasoning about Scenario agent tests (@langwatch/scenario): turn-plans/scripts, the judge, simulation runs, and "what can the judge see?" / "why does this message sequence look wrong?" questions.langwatch-expert— the LangWatch observability/evaluation platform: instrumentation, tracing, the trace schema, analytics, and reading what actually happened in a run.
The agents prefer live docs (and, with a key, live traces) over stale priors. Documentation tools work with NO API key — install is zero-config. Set the optional LangWatch API key at enable time to unlock live traces and observability (search_traces, get_trace, get_analytics, and the platform_* tools).
Found a gap in an expert's knowledge? Open an issue — contributions are welcome but never required.
Automation / CI
The interactive install prompt is the normal way to set your key — paste it once and it's stored; change it any time by re-running the plugin config. The MCP server reads it from there on every launch.
In a headless/scripted run (claude --plugin-dir ... -p ...) there is no prompt, so pass the key directly with --mcp-config instead:
{
"mcpServers": {
"langwatch": {
"command": "npx",
"args": ["-y", "@langwatch/mcp-server"],
"env": { "LANGWATCH_API_KEY": "$LANGWATCH_API_KEY" }
}
}
}
Docs tools (fetch_scenario_docs, fetch_langwatch_docs) need no key in either mode.
Related tools
@langwatch/mcp-server— the LangWatch MCP server these agents call for docs, traces, and platform access.langwatch/better-agents— a separate toolkit for building agents.langwatch/skills— LangWatch's skills-CLI capabilities (a separate distribution channel).