Production-grade practice of the open-source waveform-debugging MCP
Step-by-step: obtain wave-mcp from GitHub or PyPI, install or clone the package, provide FST/VCD waveform files and RTL netlists, start the wave-mcp service, and use its 27 MCP tools for hierarchy exploration, signal/value queries and driver analysis. Also describes integration points for Code Agents (Claude Code, Cursor, VS Code Copilot) and what to verify in the repository (license, tests, config examples).
Image Analysis
Slide 1 — Observed facts (only readable text and directly visible items recorded)
- Readable text (top to bottom, line-by-line, preserving case and punctuation):
1. AI 读波形不用买
2. License: 腾讯
3. 开源 wave-mcp
4. 波形调试 MCP 的
5. 生产级实践
- Visible UI/layout elements (facts):
- Background is light pink.
- Font color is dark gray/brown; the title line is visually larger than other lines.
- The word “License:” is in English, followed by the Chinese word “腾讯”.
- “wave-mcp” is lowercase with a hyphen; “MCP” is uppercase.
- The words “腾讯” and “波形调试” each have a cyan/light-blue highlight (rectangular background or underline-style highlight) for emphasis.
- There is a large left-quote decorative element at the top-left (light pink, darker/less transparent than the background).
- Near the bottom-right edge there is a very small pink rectangle or short dash decorative element (lighter than the text).
- Text is left-aligned with a consistent left margin.
- No tables, charts, buttons, or identifiable UI step indicators are visible.
Slide 1 — Unclear or indiscernible items
- Unable to determine if there is very small text or an icon above the small pink mark in the bottom-right (too small and close in color to the background).
- Unable to confirm whether the left-quote decoration at the top contains any fine text or graphics (appears to be a decoration).
Slide 1 — Inferences (marked as inference, not fact)
- The layout may be a slide cover or a promotional panel (based on common title/subtitle/decorative layout), but this is speculative.
- The text intent appears to introduce the open-source project “wave-mcp” and its role in “waveform debugging / production practice”, with “License: 腾讯” noted; this interpretation is speculative.
- The cyan highlights likely emphasize “腾讯” and “波形调试” (common design usage), speculative.
Notes
- The image does not contain recognizable interactive steps, UI buttons, or chart data; no timestamp, version number, or author byline was identified (if small text exists, it was marked as indiscernible).
SkillPI Creation Analysis
Purpose and scope:
- The source announces wave-mcp, an open-source MCP Server for waveform debugging from Tencent Penglai Lab's verification team. It is intended to read simulator-produced FST/VCD files plus RTL netlists and expose a set of MCP tools for waveform-based analysis (hierarchy exploration, signal/value queries, driver analysis, tracing, and static checks).
Image evidence summary:
- The promotional slide contains clear textual claims: “AI 读波形不用买”, “License: 腾讯”, “开源 wave-mcp”, indicating an announcement/marketing visual rather than an in-depth usage screenshot. No runnable UI or configuration sample is present in the image.
Verified tools/skills and practical steps (based on the provided summary):
1) Required artifacts: FST or VCD waveform files and the corresponding RTL netlist/source. wave-mcp is described as consuming these outputs rather than running simulators itself.
2) Obtain the software: repository and package locations are provided — GitHub: https://github.com/Tencent/wave-mcp and PyPI: https://pypi.org/project/wave-mcp. Use pip install or clone the repository per the project README.
3) Service deployment: run a wave-mcp server process that loads the waveform files and netlist. The summary does not include exact CLI commands or config files; consult the repository for details.
4) Tool usage: after loading data, call the provided MCP tools (27 tools across 8 categories) for hierarchy exploration, signal/value queries, driver analysis, value/X-state tracing, file/declaration queries, and static analyses.
5) Editor/agent integration: the project notes that Code Agents (Claude Code, Cursor, VS Code Copilot) can be configured with an mcpServers entry to access the server; configuration examples are not included in the summary.
Reproducibility prerequisites:
- FST/VCD waveform files and matching RTL netlists.
- A Python runtime/environment and permissions to run the wave-mcp server.
- Access to the project repository and its documentation for exact commands, configuration formats, and examples.
Items requiring in-repo verification (claims to check before relying on them):
- License and usage limits: the summary states MIT license and “no machine or concurrency limits”; verify the LICENSE file and any additional project notes in the repository for restrictions or clarifications.
- Test and correctness claims: the summary reports validation results (dozens of modules, 2.25 million signal-level validations, 100% value-query correctness, >350k tool calls across test cases). Verify test artifacts, test scripts, and result logs in the repository or its test suite to substantiate these claims.
- Integration examples: the mention of adding an mcpServers configuration for specific Code Agents requires finding concrete configuration examples in the project docs or issues.
Deliverables you can expect (based on the source):
- An open-source wave-mcp Server package (source on GitHub, PyPI distribution) capable of reading FST/VCD plus netlists and exposing MCP tool interfaces.
- A set of tools for waveform-driven verification tasks (27 tools in 8 categories) as described; usage and API details are in the project documentation.
Closing notes:
- The provided content is a summary; detailed operational commands, sample configs, and test artifacts are not included in the summary. To deploy and validate wave-mcp in your environment, consult the repository at https://github.com/Tencent/wave-mcp and the PyPI page https://pypi.org/project/wave-mcp for README, examples, and test resources. Claims about unlimited concurrency and test pass rates should be confirmed against the repository's LICENSE and test reports before assuming them in production.
