{
  "slug": "sequential-thinking-mcp",
  "name": "Sequential Thinking MCP",
  "url": "https://github.com/modelcontextprotocol/servers/tree/main/src/sequentialthinking",
  "cat": "mcp",
  "group": "ai",
  "tagline": "Reference MCP server (reasoning)",
  "price": "Free, MIT",
  "tier": "free",
  "pick": false,
  "what": "One of only seven reference servers still maintained by the MCP steering group. It gives the model a scratchpad for structured, revisable step-by-step reasoning: thoughts can be revised, branched into alternative paths, and the total step count adjusted mid-problem as the real scope becomes clear.",
  "why": "Forces a plan before code on gnarly debugging and architecture calls, where a model's instinct is to start editing files immediately. Cheap to add, and it only fires when the agent chooses to think.",
  "warn": "Pure overhead on simple tasks — every thought is another round trip and more tokens. Current reasoning models already do much of this internally, so the gain is smaller than it was in 2024.",
  "install": "claude mcp add sequential-thinking npx -y @modelcontextprotocol/server-sequential-thinking",
  "category": {
    "key": "mcp",
    "name": "MCP servers & registries",
    "desc": "How your agent reaches past the terminal. Directory scale is not a quality signal.",
    "group": "ai",
    "groupName": "Put AI inside it"
  },
  "freshness": null,
  "health": {
    "ok": true,
    "status_code": 200,
    "final_url": null,
    "checked_at": "2026-08-19T00:17:14.278+00:00"
  },
  "corrected_at": null,
  "upstream_changed_at": null
}