board-spec-mcp
MCP server that serves Qualcomm Yocto board specs from a git repo, enabling retrieval and validation of board specifications.
README
board-spec-mcp
MCP server that serves Qualcomm Yocto board specs (machine_creation +
partition_conf fields) from the roshs189/board-spec
git repo — one branch per machine. Built to replace reading a Confluence
"Partition Spec" page in the qcom-yocto-new-machine and
qcom-partition-conf-new-board Claude Code skills.
What it does
- Clones/fetches
board-specinto a local cache (~/.cache/board-spec-mcp/repoby default). - Reads a board's spec straight from its branch (
origin/<machine>:boards/<machine>/spec.yaml) without checking it out — no working-tree conflicts, always in sync with arefresh_specscall. - Validates specs against
schema/board-spec.schema.json(onmain) plus the cross-checks the skills used to do by hand:delta=newpartitions need anotesjustification and must not collide with SoC-mandated names (xbl/tz/hyp/aop/...),partition_files_subdirmust equalpartitions/<machine>/<storage_type>, chip IDs shouldn't collide across boards.
Setup (anyone, from scratch)
Requires Python 3.10+ and an SSH key with read access to
github.com/roshs189/board-spec (or set BOARD_SPEC_REPO_URL to a repo you
do have access to — see Configuration below).
git clone <this-repo-url> board-spec-mcp
cd board-spec-mcp
python3 -m venv .venv
.venv/bin/pip install -e .
Verify it starts cleanly:
.venv/bin/board-spec-mcp # should hang waiting on stdin — Ctrl-C to exit
Register with Claude Code
Add a .mcp.json in the project you want this available in (or merge into
an existing one):
{
"mcpServers": {
"board-spec": {
"command": "/absolute/path/to/board-spec-mcp/.venv/bin/board-spec-mcp"
}
}
}
Restart Claude Code (or run /mcp to reconnect) and the tools below become
available.
Configuration
Environment variables, all optional:
| Variable | Default | Purpose |
|---|---|---|
BOARD_SPEC_REPO_URL |
git@github.com:roshs189/board-spec.git |
Remote to clone/fetch |
BOARD_SPEC_CACHE_DIR |
~/.cache/board-spec-mcp/repo |
Local bare-ish clone location |
Set these in the .mcp.json entry's "env" block if you need a fork or a
different cache path:
{
"mcpServers": {
"board-spec": {
"command": "/absolute/path/to/board-spec-mcp/.venv/bin/board-spec-mcp",
"env": { "BOARD_SPEC_REPO_URL": "git@github.com:<you>/board-spec.git" }
}
}
}
Tools
list_boards()— every machine with a branch inboard-spec(excludesmain).get_board_spec(machine)— full spec.get_machine_creation_fields(machine)— just themachine_creationsection.get_partition_conf_fields(machine)— just thepartition_confsection.validate_board_spec(machine)—{valid, schema_errors, warnings}.refresh_specs()— fetch latest branches/commits (call after a spec is pushed).
Also exposed as a resource: board-spec://<machine>.
Adding a new board
Specs are human-authored and PR-reviewed in board-spec itself — this server
only reads, never writes. See that repo's README for the branch/PR workflow.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。