GetSkillary

MCP Registry

Use the GetSkillary MCP registry

Local AI agents can connect to the GetSkillary MCP endpoint to search public skills, inspect details, request related packages, retrieve manual download links, map natural-language scenarios to solution workflows, and record aggregate usage events.

Updated: 2026-07-09

Direct Answer

Use the GetSkillary MCP registry when Codex receives a natural-language scenario and needs a structured route to a GetSkillary solution, skill bundle, manual install plan, and local verification steps. The solution-first workflow is search_solutions -> get_solution_detail -> recommend_solution_skills -> get_solution_install_plan.

Canonical endpoint: https://mcp.getskillary.com/mcp. The skill-first workflow remains search_skills -> get_skill_detail -> get_download_url -> get_install_guide when the user already knows the exact package category or skill they want to inspect.

Solution-First Agent Flow

1. Start from the scenario

Ask Codex in plain language, such as I need Codex to validate a broken JSON config before editing anything.

2. Match a solution

Codex calls search_solutions, then get_solution_detail for the best matching solution slug.

3. Expose the bundle

Codex calls recommend_solution_skills so the user can inspect the primary and supporting skills.

4. Return an install plan

Codex calls get_solution_install_plan for manual review, local setup, and verification steps.

5. Ask for confirmation

Codex should not install, execute, collect credentials, or deploy remotely without a separate user-approved local action.

Skill-First Agent Flow

1. Search by task

Ask the agent to search for the work you need done, such as markdown writing workflow.

2. Inspect details

Have the agent compare summaries, use cases, package size, tags, and manual download links.

3. Retrieve download URL

Use get_download_url only to return the manual zip link for the selected package.

4. Download manually

Choose a package yourself, read SKILL.md, then install it in your local skills directory.

Endpoint

https://mcp.getskillary.com/mcp

Client configuration

Add the remote server URL in an MCP-compatible client that supports HTTP MCP servers.

{
  "mcpServers": {
    "getskillary": {
      "url": "https://mcp.getskillary.com/mcp"
    }
  }
}

Available tools

  • search_solutions: search scenario solution records by user need or task language.
  • get_solution_detail: inspect a solution page record, answer fields, risk boundary, and target URL.
  • recommend_solution_skills: return the primary and supporting skill bundle for a solution.
  • get_solution_install_plan: return manual install and verification guidance for the solution bundle.
  • search_skills: search public skill records by query, category, or tag.
  • get_skill_detail: inspect one public skill record before downloading.
  • recommend_skills: find related skills by slug, query, or category.
  • get_download_url: return the manual zip download URL for a public skill.
  • get_install_guide: return manual installation guidance for a selected skill.
  • record_usage: record an aggregate usage event after the user chooses to proceed.

Example Solution Request

{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "tools/call",
  "params": {
    "name": "search_solutions",
    "arguments": {
      "query": "I need Codex to turn rough requirements into a build plan",
      "limit": 5
    }
  }
}

Example Skill Request

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "search_skills",
    "arguments": {
      "query": "markdown writing workflow",
      "limit": 5
    }
  }
}

Operation boundary

The MCP registry discovers, recommends, and plans. Codex executes locally only after user confirmation. The MCP registry does not install files, modify local configuration, expose quarantined skills, publish private analytics views, handle credentials, or perform remote deployment.

Do not send API keys, browser session data, private customer records, external account access, or production authority through the MCP response. Use local verification commands, smoke tests, or review checklists before accepting output.

Discovery files

AI crawlers and local tooling can also use llms.txt, llms-full.txt, skills.json, and mcp.json to understand the public catalog without reading internal review data.

Browse skills Manual install guide