VPS Hardening
Server Details
Read-only security checklists: VPS hardening, Docker, secrets, Supabase, Stripe, auth, AI agents.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- Repository
- NoteSupreme/vps-hardening-mcp
- GitHub Stars
- 0
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 5 tools
The five tools split cleanly into search, enumerate, and read, and the descriptions explicitly rank and cross-reference each other (digest first, list for flat enumeration, overview for authored prose). The one soft boundary is get_blueprint_overview versus get_ship_digest, which share a fallback path and could be confused despite the prose explaining the difference.
All names are snake_case verb_noun (get_*, list_*, read_node, search_nodes) and read predictably. The 'ship' scoping is applied inconsistently (get_ship_digest, list_ship_nodes vs read_node, search_nodes), which is a minor wart rather than a real obstacle.
Five tools is a tight, well-scoped surface for a read-only navigation/retrieval server, with no filler. Each tool covers a distinct access mode: orient, enumerate, search, read, and authored overview.
The retrieval lifecycle is covered end to end — orient, browse sections, flat list, full-text/typed search, and paginated body reads with per-heading slicing. The main gap is graph traversal: outbound_links are returned, but there is no inbound/backlinks tool, so discovery of reverse references is a dead end.
Available Tools
5 toolsvpshardening__get_blueprint_overviewAInspect
Read the human-AUTHORED overview prose for the Ship (omit node_id) or for one section. Useful only where someone has written a summary; it is not the way to navigate — vpshardening__get_ship_digest is the map, call that first. When no summary has been written this falls back to the same section map vpshardening__get_ship_digest returns, drilled with { section_id } and paged with { cursor }.
Inputs: cursor, node_id, section_id.
vpshardening__get_ship_digestAInspect
Orient on this Ship — ALWAYS call this first. With no arguments it returns the section MAP and nothing else: every top-level section with its description and child_count, which is small and cheap even on a Ship of hundreds of nodes. Then pass { section_id } to list one section's direct children, and vpshardening__read_node for the ones worth reading. A large section arrives in pages — continue it with { cursor }.
Inputs: cursor, section_id.
vpshardening__list_ship_nodesAInspect
List every node: node_id, title, parent_id, node_type, plus type/description/tags/resource when set. No bodies. This is a flat list of the WHOLE Ship — on a large one it is long and arrives in pages, so pass { cursor } to continue. To navigate rather than enumerate, use vpshardening__get_ship_digest.
Inputs: cursor.
vpshardening__read_nodeAInspect
Read one node's full markdown body + frontmatter by node_id. Response includes outbound_links[] (resolved graph edges) unless with_links:false, plus source attribution. Pass heading (e.g. "Tactics") to get ONLY that markdown section — a few hundred chars instead of the whole body; every heading read, hit or miss, carries the node's headings list, and a miss is heading_not_found — so a guessed heading is cheap (heading "" just lists the headings). A heading read is partial — never write it back as a node's content; re-read without heading first. A long body is returned one page at a time — when the response says truncated, call again with { cursor } to continue it.
Inputs: node_id (required), cursor, heading, with_links.
vpshardening__search_nodesAInspect
Search nodes by text, node_type, and/or semantic type. Token-AND across title + description + tags. Primary retrieval path — use it before reading bodies. Results are ranked best match first — score on each row is the match strength (title hits count double). Page 1 is the top of the ranking. A one-word query can match hundreds of rows — put two or more terms in query so the top rows are the answer, not a dump. Terms are AND-ed: if two return nothing, drop one. Reports total_matches; when more matches exist than fit in one response, narrow the query or pass { cursor } for the next page (a cursor is only valid for the query that produced it).
Inputs: type, query, cursor, node_type.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Supabase RLS simulator, app key-leak check, email sender check, DMARC and payment-fraud checks.
Security checks for AI-built apps (Lovable, Supabase, Bolt): RLS, headers, monitoring, fixes.
Security checks for Next.js + Supabase apps: code scan, read-only live DB check, fix confirmation.
Security scans for AI-built apps: leaked keys, open Supabase RLS, vulnerable packages, live sites.
Related MCP Servers
AlicenseAqualityCmaintenanceServer security auditing (413 checks, 29 categories), production hardening, and fleet management. Supports Hetzner, DigitalOcean, Vultr, and Linode.17626 npm59Apache 2.0- AlicenseAqualityCmaintenanceEnables AI coding assistants to check whether an API key is safe to expose, audit Supabase Row Level Security, generate owner-based RLS policy templates, and run a builder-specific pre-launch security checklist. Runs locally over stdio with no network calls, telemetry, or file access.5MIT
- AlicenseAqualityBmaintenanceRead-only Supabase Row Level Security review. Analyzes an exported catalog query (no keys needed) for RLS turned off, policies that are just true, anon-callable SECURITY DEFINER functions, views exposing auth.users, public Storage buckets and more, and explains each rule.5914 PyPIMIT
- AlicenseAqualityDmaintenanceAudits PostgreSQL/Supabase schemas for security issues like missing RLS, permissive policies, and sensitive data exposure during AI conversations.3MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.