flixbridge-mcp

name	flixbridge-mcp
description	Use when operating, validating, or troubleshooting the FlixBridge Arr MCP server, especially Sonarr/Radarr/SABnzbd queue, import, downloader, diagnostics, safe production mutation, and agent-browser UI validation workflows.

FlixBridge MCP

Required Order

• Call list_services first and use only returned service/downloader names.
• Treat all other tool calls as scoped to one returned service unless the tool explicitly accepts multiple services.
• Keep outputs in the stable { ok, data?, error? } contract; do not infer success from missing errors alone.

Read-Only Baseline

• Run system_status for each target Arr service.
• Run queue_list with a bounded pageSize.
• Run import_issues for missing/import problem context.
• Run queue_diagnostics with autoFix:false for production baselines.
• Run download_status when a downloader is configured.
• Run server_metrics before and after validation work.

Mutation Rules

• Assume queue_diagnostics and all_services_diagnostics mutate unless autoFix:false is explicitly set.
• Prefer remove_content with dryRun:true before any queue, library, or downloader removal.
• Execute remove_content only with the same ids/options and the returned confirmationToken.
• Limit production mutations to reviewed Sample, not-an-upgrade, explicit retry, or unambiguous manual-import cases.
• Do not delete downloader data with deleteFiles:true unless explicitly approved for that item or batch.

Stuck Queue Triage

• For Sample items, require matching MCP and UI evidence by service, queue id, title/path, download id, status, and reason.
• For not-an-upgrade items, require diagnostics text such as not-a-custom-format-upgrade or does-not-improve-existing before queue removal.
• For Found matching series via grab history...Automatic import is not possible, attempt manual import only when the candidate maps to exactly one matching series/movie id and the UI agrees.
• If MCP output and UI disagree on target identity, stop and ask before mutating.

Browser Validation

• Use the agent-browser skill for Sonarr, Radarr, and SABnzbd UIs only.
• Before starting, run agent-browser --version and agent-browser session list; record pre-existing sessions.
• Use a new named session for validation, take snapshots/screenshots before and after mutation, then close sessions created during the task.
• Use the UI for evidence and verification; prefer MCP tools as the primary mutation path.

Evidence To Capture

• Service/downloader names from list_services.
• Tool input and structured output with secrets omitted.
• Queue item id, title, output path, download id, status, reason, and proposed action.
• UI screenshot or extracted UI text proving the same target.
• Before/after queue_list, queue_diagnostics autoFix:false, download_status, and server_metrics summaries.
• Recovery path for each approved mutation, usually re-search or re-grab.

Local Verification

• Run npm run check after docs or code edits.
• Run npm run lint, npm test, and npm run build for behavior changes.
• Run npm run smoke only when environment variables point at the intended instance and no mutating defaults are enabled.