generated: '2026-08-30'
method: searched
source: https://github.com/amd/ryzenai-mcp-server (AMD's own GitHub organization)
name: AMD Ryzen AI MCP Server
description: >-
  AMD publishes a first-party Model Context Protocol server, amd/ryzenai-mcp-server, in its own
  GitHub organization. It gives an MCP client semantic search over the AMD Ryzen AI documentation
  (https://ryzenai.docs.amd.com) and the amd/RyzenAI-SW code samples repository, backed by a local
  embedding index. It is a documentation/knowledge server, not a control plane for AMD hardware or
  cloud resources.
status: published
deployment:
  mode: local-stdio
  endpoint: null
  install: "git clone https://github.com/amd/ryzenai-mcp-server && cd ryzenai-mcp-server && pip install -r requirements.txt && python server.py"
  package: https://github.com/amd/ryzenai-mcp-server
  auth: none
  verified: searched
  note: >-
    AMD ships NO hosted/remote MCP endpoint. The repository's own server.json declares
    `"transport": ["stdio"]` and `"config": {"command": "python", "args": ["server.py"]}`, and
    .mcp.json registers it as a stdio server with `cwd: ${PLUGIN_ROOT}`. There is no npm or PyPI
    distribution — installation is a git clone plus `pip install -r requirements.txt`. A human must
    install and run it on a machine before any agent can reach it. GITHUB_TOKEN is an OPTIONAL
    environment variable used only to raise GitHub API rate limits while indexing; the server itself
    requires no credential.
repository: https://github.com/amd/ryzenai-mcp-server
license: MIT
language: Python
created: '2026-01-27'
last_pushed: '2026-04-08'
runtime:
  requires: Python 3.8+
  key_dependencies:
    - mcp>=1.14.0
    - sentence-transformers>=2.2.0
    - chromadb>=0.4.0
    - langchain-community>=0.0.20
    - tree-sitter-languages>=1.10.0
  environment:
    - name: GITHUB_TOKEN
      required: false
      purpose: Higher GitHub API rate limits while indexing; requires public_repo scope.
    - name: RYZENAI_INDEX_REFRESH_HOURS
      required: false
      default: 24
      purpose: Hours between automatic index refreshes; 0 disables auto-refresh.
    - name: RYZENAI_INDEX_MAX_FILES
      required: false
      default: 100
      purpose: Maximum files downloaded per refresh.
tools:
  - name: read_ryzenai
    description: Read AMD Ryzen AI documentation pages.
  - name: search_ryzenai_sw_keyword
    description: Keyword search in the Ryzen AI repository.
  - name: search_ryzenai_sw_enhanced
    description: Enhanced semantic search with AI embeddings.
  - name: read_ryzenai_sw
    description: Read specific files from the repository.
  - name: list_ryzenai_sw_directory
    description: List contents of a directory.
  - name: find_ryzenai_sw_files
    description: Find files by name pattern or type.
  - name: download_and_index_ryzenai_enhanced
    description: Download and index repository with AI embeddings.
  - name: get_enhanced_embedding_status
    description: Check embedding index status.
  - name: refresh_ryzenai_index
    description: Clear the index and re-download/re-index after repo or docs updates.
tool_count: 9
tool_schemas:
  captured: false
  note: >-
    Tool names and descriptions are transcribed verbatim from the repository README and server.json.
    Full inputSchema definitions were NOT captured: there is no remote endpoint to POST
    tools/list against, and reading them would require cloning and running the stdio server locally.
    Recorded honestly rather than guessed.
evidence:
  - url: https://github.com/amd/ryzenai-mcp-server
    status: 200
  - url: https://api.github.com/repos/amd/ryzenai-mcp-server/contents/server.json
    status: 200
  - url: https://api.github.com/repos/amd/ryzenai-mcp-server/readme
    status: 200
related:
  - https://ryzenai.docs.amd.com
  - https://github.com/amd/RyzenAI-SW