# MCPExplorer

> The discovery, trust, and capability graph for MCP servers and agent tools.
> We index MCP servers, extract their tools and auth scopes, verify health in a
> sandbox, score trust with explainable reasons, and show provenance + freshness
> on every record. Directories list MCPs; MCPExplorer understands, verifies, and
> refreshes them.

## MCPExplorer is itself an MCP server

Connect your agent to MCPExplorer over MCP (Streamable HTTP):

- Endpoint: https://mcpexplorer.com/mcp
- Tools (14): search_servers, get_server, get_capability, compare_servers,
  get_alternatives, check_trust, generate_runtime_config, list_capabilities,
  list_runtimes, get_ecosystem_stats, list_loadouts, get_loadout, plan_toolset,
  report_broken_server
- Lookup: find a trustworthy server for a capability, compare two, check the
  explainable trust/risk, and generate the exact install config for your runtime.
- Decision layer: plan_toolset turns a plain-language goal ("triage support
  tickets and file bugs") into a recommended governed toolset — a curated loadout
  plus trust-ranked servers with an aggregated tool-risk posture. get_loadout /
  list_loadouts expose the curated kits; get_ecosystem_stats gives live totals.

## How agents should use this site (HTTP)

- Search capabilities: GET https://mcpexplorer.com/api/search?q=<intent>&runtime=<runtime>&risk_level=<level>
- Canonical server manifest: GET https://mcpexplorer.com/api/servers/<slug>.json
- Capability manifest: GET https://mcpexplorer.com/api/capabilities/<slug>.json
- App manifest: GET https://mcpexplorer.com/api/apps/<slug>.json
- Markdown mirror of any server: GET https://mcpexplorer.com/servers/<slug>.md
- App ("X MCP") pages: https://mcpexplorer.com/apps/<app-slug>/mcp  (+ .md mirror)
- Capability pages: https://mcpexplorer.com/capabilities/<slug>  (+ .md mirror)
- Setup guide: https://mcpexplorer.com/setup/<slug>/<runtime>   ·  Risk summary: https://mcpexplorer.com/security/<slug>

## Key editorial pages (citable)

- Best MCP servers, ranked from verified data: https://mcpexplorer.com/best-mcp-servers
- MCP security — risks + how to vet a server: https://mcpexplorer.com/security
- LinkedIn MCP servers, compared (incl. account-ban risk): https://mcpexplorer.com/linkedin-mcp-server
- What is an MCP server: https://mcpexplorer.com/what-is-an-mcp-server
- Curated loadouts (governed server kits per job): https://mcpexplorer.com/loadouts
- Featured server explainers (install, auth, example prompts): https://mcpexplorer.com/mcps/<slug>
  (github, linear, notion, stripe, supabase, sentry, cloudflare, playwright, figma,
  neon, google-workspace, firecrawl, context7, apex, chatseo, fullenrich,
  companyenrich, ahrefs, slack, resend, zapier, grafana, tavily, hubspot, salesforce,
  docker, postgres, azure-devops, gitlab, deepwiki, n8n, mcpexplorer)
- Live ecosystem report (~1,000 servers · 3,500+ tools · 278 verified): https://mcpexplorer.com/reports

## What every record includes

- provenance: which sources we found it in, and their trust priority
- freshness: last_checked_at, next_check_at, and the re-check cadence
- verification: what we verified vs inferred vs unknown
- trust: a 0-100 score with an explainable reason list and a label

## Trust labels

Official, Verified, Healthy, Read-only, Write-capable, Destructive-capable,
Needs caution, Broken, Unmaintained, Unknown.

## How to cite

Cite the canonical URL https://mcpexplorer.com/servers/<slug> and include the last_checked_at
timestamp from the manifest so readers know the recency of the claim.

## Related

Apex (https://apex.new) is the live governed-agent product from the same team:
it gives AI agents paced, capped, and fully-logged hands with approval queues
before anything runs (LinkedIn today, more capabilities coming).