Vivado MCP
A Model Context Protocol server that lets AI clients operate AMD Vivado through safe, workflow-level tools while allowing the user to watch and interact with the Vivado GUI.
README
Vivado MCP
Vivado MCP is planned as a Model Context Protocol server that lets AI clients operate AMD Vivado through safe, workflow-level tools while the user can watch and interact with the Vivado GUI.
The first design target is not GUI click automation. The preferred mode is a managed Vivado Tcl session that can open the GUI with start_gui, load a small Tcl bridge, and let the MCP server submit validated workflow commands into that same Vivado process. Batch mode remains useful for CI and fallback automation.
Current design documents:
- Vivado MCP Design
- ADR 0001: Control Vivado through Tcl batch mode
- ADR 0002: Support managed GUI Tcl sessions
- ADR 0003: Support trusted-local raw Tcl
- ADR 0004: Provide built-in help and skills
- Built-in Skills
Initial scope
- Discover a local Vivado installation and report its version.
- Start and stop a managed Vivado Tcl/GUI session.
- Submit raw Tcl to a managed session when trusted-local expert mode is enabled.
- Create or open project-mode Vivado projects.
- Add RTL/source/constraint files with path validation.
- Run synthesis, implementation, and bitstream generation.
- Generate timing, utilization, DRC, and message reports.
- Parse common report outputs into compact structured summaries.
- Expose logs and generated reports as MCP resources.
- Provide built-in help/skills so AI clients can learn the intended Vivado workflows before acting.
Capability profiles
safe: workflow tools only; no raw Tcl.trusted-local: workflow tools plus raw Tcl/source-file execution inside the managed Vivado session.unrestricted: raw Tcl with minimal policy checks for personal local use.
The prototype bridge in experiments/mcp_bridge.tcl proves the core control path: an external process can submit Tcl files to a live Vivado Tcl/GUI session and receive result files back.
Built-in help
The MCP should expose tutorial content through both tools and resources:
vivado_helpvivado_list_skillsvivado_get_skillvivado_suggest_next_stepsvivado://skills/index
Seed skill docs live in docs/skills.
Install For Local Use
From this repo:
python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -e ".[dev]"
This machine has been tested with:
C:\Xilinx\Vivado\2023.1\bin\vivado.bat
MCP Client Configuration
Use the installed console script as the MCP server command:
{
"mcpServers": {
"vivado": {
"command": "C:\\Workspace\\Vivado_mcp\\.venv\\Scripts\\vivado-mcp.exe",
"env": {
"VIVADO_BIN": "C:\\Xilinx\\Vivado\\2023.1\\bin\\vivado.bat",
"VIVADO_MCP_WORKSPACE": "C:\\Workspace\\Vivado_mcp",
"VIVADO_MCP_ALLOWED_ROOTS": "C:\\Workspace\\Vivado_mcp"
}
}
}
}
VIVADO_MCP_WORKSPACE is the default working directory for managed sessions. VIVADO_MCP_ALLOWED_ROOTS is a semicolon-separated list on Windows; workflow paths such as projects, sources, constraints, and Tcl files in trusted-local mode must stay under one of these roots. Use unrestricted capability profile only for personal experiments that need to source Tcl outside the allowed roots.
First Manual Test
After connecting the MCP client, use this sequence:
vivado_helpwithtopic="gui_session".vivado_check_installation.vivado_start_sessionwithopen_gui=true.vivado_run_tclwithtcl="return \"version=[version -short]\"".vivado_project_summaryafter opening or creating a project.vivado_list_artifactsto inspect command/result files.vivado_stop_session.
Implemented Tools
vivado_check_installationvivado_start_sessionvivado_list_sessionsvivado_session_statevivado_stop_sessionvivado_run_tclvivado_source_tclvivado_create_projectvivado_open_projectvivado_add_sourcesvivado_run_synthesisvivado_run_implementationvivado_generate_bitstreamvivado_reportvivado_project_summaryvivado_list_artifactsvivado_read_artifactvivado_helpvivado_list_skillsvivado_get_skillvivado_suggest_next_steps
Development Checks
.\.venv\Scripts\python.exe -m pytest
.\.venv\Scripts\python.exe -m compileall src
The test suite includes a fake Vivado process and an MCP protocol smoke test that starts the stdio server and lists tools/resources.
Artifact Resources
Command files, result files, logs, and reports are stored under the managed session directory and exposed through artifact URIs:
vivado://sessions/{session_ref}/artifacts/{artifact_id}
Use vivado_list_artifacts to discover artifact URIs and vivado_read_artifact to read text artifacts. vivado_report also returns a best-effort report_summary for timing, utilization, DRC, and message reports. vivado_project_summary returns the current project, source files, runs, IP, and block designs as structured data.
Explicitly out of scope for the first version
- GUI click automation.
- Attaching to an arbitrary already-open Vivado process that did not load the MCP bridge.
- Hardware programming and live hardware manager operations.
- Full IP Integrator automation.
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。