consultando-sentry

name

consultando-sentry

description

Consulta issues, eventos e metricas do Sentry via MCP Server. Use quando o usuario mencionar "Sentry", "issues do Sentry", "erros em producao", "bugs no Sentry", "exceptions nao tratadas", "500 errors no Sentry", "resolver issue", "marcar resolvido no Sentry", ou qualquer variacao que envolva monitoramento de erros. Tambem usar quando o usuario pedir para "ver erros", "checar exceptions", "quantos bugs tem", "issues abertas", "erros das ultimas 24h", ou "root cause analysis". NAO usar para logs do Render (usar MCP Render list_logs), metricas de CPU/memoria (usar MCP Render get_metrics), ou diagnostico de banco (usar diagnosticando-banco).

Consultando Sentry

Skill para consultar e operar o Sentry da Nacom Goya via MCP Server (acesso direto, sem script intermediario).

Contexto

• Organizacao: nacom
• Projeto: python-flask (unico)
• Region URL: https://us.sentry.io (OBRIGATORIO em todas as tools)
• Auth: OAuth via MCP Server (Bearer token configurado em ~/.claude/settings.json)
• Dashboard Web: https://nacom.sentry.io/
• Referencia: .claude/references/INFRAESTRUTURA.md secao "Sentry"

Como Usar — MCP Tools (metodo primario)

1. Listar issues nao resolvidas

mcp__sentry__search_issues(
    organizationSlug="nacom",
    projectSlugOrId="python-flask",
    naturalLanguageQuery="unresolved issues from last 24 hours",
    regionUrl="https://us.sentry.io",
    limit=25
)

Exemplos de queries naturais:

• "critical bugs from last week"
• "unhandled errors affecting 100+ users"
• "issues assigned to me"
• "errors in devolucao module"
• "user feedback from production"

2. Eventos / stacktrace de uma issue

NAO existe get_issue_details neste MCP server. Para stacktrace e detalhes dos eventos de uma issue use search_issue_events; para causa raiz + arquivos afetados use analyze_issue_with_seer (secao 3).

mcp__sentry__search_issue_events(
    organizationSlug="nacom",
    issueId="PYTHON-FLASK-5",
    naturalLanguageQuery="latest event with stacktrace",
    regionUrl="https://us.sentry.io"
)

3. Root Cause Analysis com Seer (AI)

mcp__sentry__analyze_issue_with_seer(
    organizationSlug="nacom",
    issueId="PYTHON-FLASK-5",
    regionUrl="https://us.sentry.io"
)

Retorna: causa raiz, arquivos/linhas afetados, sugestoes de fix com codigo. Resultados cached (~2-5min primeira vez).

4. Resolver/Ignorar issues

mcp__sentry__update_issue(
    organizationSlug="nacom",
    issueId="PYTHON-FLASK-5",
    status="resolved",
    regionUrl="https://us.sentry.io"
)

Status validos: resolved, resolvedInNextRelease, unresolved, ignored.

5. Eventos de uma issue (filtrados)

mcp__sentry__search_issue_events(
    organizationSlug="nacom",
    issueId="PYTHON-FLASK-5",
    naturalLanguageQuery="from last hour in production",
    regionUrl="https://us.sentry.io"
)

6. Contagens e estatisticas

mcp__sentry__search_events(
    organizationSlug="nacom",
    projectSlug="python-flask",
    naturalLanguageQuery="how many errors today",
    regionUrl="https://us.sentry.io"
)

Exemplos: "count of database failures this week", "total tokens used by model", "average response time".

7. Tags de uma issue (distribuicao)

mcp__sentry__get_issue_tag_values(
    organizationSlug="nacom",
    issueId="PYTHON-FLASK-5",
    tagKey="url",
    regionUrl="https://us.sentry.io"
)

Tag keys comuns: url, browser, environment, release, user, transaction.

8. Releases

mcp__sentry__find_releases(
    organizationSlug="nacom",
    projectSlug="python-flask",
    regionUrl="https://us.sentry.io"
)

9. Documentacao Sentry

NAO existe get_trace_details neste MCP server. Para consultar a documentacao oficial do Sentry use search_docs + get_doc.

mcp__sentry__search_docs(query="distributed tracing setup", maxResults=3)

Tools MCP Disponíveis (principais)

Tool	Funcao
whoami	Identificar usuario autenticado
find_organizations	Listar organizacoes
find_projects	Listar projetos
find_teams	Listar teams
find_releases	Listar releases
find_dsns	Listar DSNs do projeto
search_issues	Buscar issues (linguagem natural)
search_events	Buscar eventos + contagens/agregacoes
search_issue_events	Filtrar eventos de uma issue (stacktrace)
get_issue_tag_values	Distribuicao de tags
get_event_attachment	Baixar attachments
analyze_issue_with_seer	Root cause analysis AI (Seer)
update_issue	Resolver/ignorar/atribuir issue
update_project	Atualizar config do projeto
create_project / create_team / create_dsn	Criar recursos
search_docs / get_doc	Documentacao Sentry

NAO existem neste MCP server: get_issue_details e get_trace_details (use search_issue_events + analyze_issue_with_seer no lugar). Tools de snapshot/profile/replay existem mas nao sao usadas por esta skill.

Metodo Secundario — Sentry CLI

Para uso no terminal (quando MCP nao esta disponivel):

sentry issue list                        # Listar issues
sentry issue explain PYTHON-FLASK-5      # Root cause analysis (Seer)
sentry issue plan PYTHON-FLASK-5         # Plano de fix automatico

Fluxo Recomendado

1. Visao geral: search_issues com "unresolved issues" ou search_events com "how many errors today"
2. Investigar: search_issue_events para stacktrace + analyze_issue_with_seer para causa raiz
3. Contexto: get_issue_tag_values para entender distribuicao (URLs, browsers, environments)
4. Corrigir: implementar fix no codigo
5. Fechar: update_issue com status "resolved"

IMPORTANTE

• SEMPRE incluir regionUrl="https://us.sentry.io" nas tools que aceitam esse parametro
• O MCP Server substitui completamente o script REST antigo (sentry_api.py)
• search_issues retorna LISTA de issues. Para contagens, usar search_events
• analyze_issue_with_seer pode demorar 2-5min na primeira vez (resultado cached depois)