inariwatch（inariwatch 搜索）

InariWatch — AI Monitoring That Fixes Your Code

You are connected to InariWatch via MCP. You have access to 25 tools for monitoring, diagnosing, and fixing production issues autonomously.

InariWatch monitors GitHub, Vercel, Netlify, Cloudflare Pages, Render, Sentry, Datadog, Expo, and custom apps via the Capture SDK. When something breaks, it diagnoses the root cause with AI, generates a fix, opens a PR, waits for CI, and auto-merges through 11 safety gates.

Available tools

Check status

• "What's breaking in production?" -> use query_alerts to list critical alerts
• "Show me error trends this week" -> use get_error_trends with days: 7
• "Is my site up?" -> use get_uptime to check all monitors
• "Show project status" -> use get_status for projects, integrations, and alert counts

Diagnose issues

• "What caused this error?" -> use get_root_cause with the alert ID
• "Assess risk for this PR" -> use assess_risk with owner, repo, and PR number
• "Has anyone else fixed this?" -> use search_community_fixes with the error message
• "Generate a post-mortem" -> use get_postmortem with the alert ID
• "Get build logs" -> use get_build_logs for the latest failed deploy

Fix issues

• "Fix the latest critical alert" -> use trigger_fix with the alert ID to start AI remediation
• "Roll back production" -> use rollback_deploy with the project ID (works on Vercel, Netlify, Cloudflare Pages, Render)
• "Silence this alert" -> use silence_alert with the alert ID
• "Acknowledge this alert" -> use acknowledge_alert to mark as read
• "Reopen this alert" -> use reopen_alert to reopen a resolved alert

Monitor and explore

• "Add uptime monitoring for api.example.com" -> use create_uptime_monitor with the URL
• "Run a health check" -> use run_health_check for all monitors
• "Check if this can be reproduced" -> use reproduce_bug with the alert ID
• "Simulate this fix" -> use simulate_fix with the alert ID and proposed changes
• "Verify the remediation worked" -> use verify_remediation with the session ID
• "Search my codebase for auth logic" -> use search_codebase with the query
• "Reindex my codebase" -> use reindex_codebase with the project ID
• "Ask about my infrastructure" -> use ask_inari with a natural language question

Important rules

1. Always confirm before destructive actions. trigger_fix and rollback_deploy modify production code and deployments. Ask the user for explicit confirmation before executing these.
2. rollback_deploy is irreversible in the sense that it changes the live deployment. The previous deployment is still available but the rollback happens immediately.
3. Rate limits apply. Query tools: 200/min. Analysis tools: 30/min. Execution tools (fix, rollback): 5/min. If you hit a limit, wait and retry.
4. All actions are scoped to the user's projects via their InariWatch token. You cannot access other users' data.
5. trigger_fix starts an async pipeline. It returns a session ID immediately. The AI remediation runs in the background: diagnose -> read code -> generate fix -> security scan -> self-review -> push -> CI -> PR -> auto-merge gates. You can check progress with verify_remediation.
6. Sampling tools (get_root_cause, assess_risk, ask_inari, simulate_fix) return context for YOU to analyze. Process the returned data and provide your own analysis to the user.

Setup

The user needs an InariWatch account at https://app.inariwatch.com and a token from Settings -> API Tokens.

Option 1 — Auto-detect (recommended):

npx @inariwatch/mcp init

This detects OpenClaw and configures automatically.

Option 2 — Manual CLI:

openclaw mcp set inariwatch '{"url":"https://mcp.inariwatch.com","transport":"streamable-http","headers":{"Authorization":"Bearer YOUR_TOKEN"}}'

Option 3 — Edit config directly: Add to ~/.openclaw/openclaw.json:

{
  "mcp": {
    "servers": {
      "inariwatch": {
        "url": "https://mcp.inariwatch.com",
        "transport": "streamable-http",
        "headers": {
          "Authorization": "Bearer YOUR_TOKEN"
        }
      }
    }
  }
}