Trigvale 的 MCP 服务器 —— 在构建代理搭建代码之前验证创业想法。
这是 Trigvale 的官方模型上下文协议 (Model Context Protocol) 服务器。它向支持 MCP 的构建代理（Cursor、Claude Code、Codex、Claude Desktop）提供了一个工具：validate_idea。预期的用户体验非常直接：您的代理在编写任何一行脚手架代码之前调用 validate_idea。绿色或琥珀色结论表示可以开始编码；红色结论则会指出您应该首先测试的假设。
validate_idea 返回的内容10 维度评分卡 — 创业准备度评分 (VRS, 0–100)，根据模型针对已发布基准的各维度评分，在代码中确定性地计算得出。模型本身从不选择 VRS 分数。
证据摘要 — 每个项目都标记了 sourceKind ∈ { observed (观察到的), inferred (推断的), missing (缺失的), ai (AI 生成的), user-claim (用户声明) }，以便您了解哪些是基于事实的，哪些是假设的。
结论 — kill (终止)、pivot (转型)、test (测试) 或 build (构建)。根据设计，build 的情况很少见。
创始人契合度校准 — 根据用户声明的技能图谱（仅限声明的输入，从不根据过去的失败进行推断）应用于各维度的增量。
1–3 个原型分配 — 将想法归类为已知模式（例如“垂直 AI SaaS”、“代理工作流”），并提供针对该类别的结构性警告。
实时证据 — 每 6 小时刷新一次，针对该想法的主要原型，从 Reddit、GitHub、Hacker News、Stack Overflow 和 dev.to 实时获取证据。
完整输出模式和方法论：https://trigvale.com/methodology
Related MCP server: product-spec-mcp
npx -y @trigvale/mcp您需要一个 Trigvale 代理令牌（Integrate 计划，$99/月）。请在 https://trigvale.com/settings 获取。
将其放入您的 MCP 客户端配置（mcpServers 块）中：
~/.cursor/mcp.json){
"mcpServers": {
"trigvale": {
"command": "npx",
"args": ["-y", "@trigvale/mcp"],
"env": {
"TRIGVALE_AGENT_TOKEN": "tvk_xxx"
}
}
}
}~/Library/Application Support/Claude/claude_desktop_config.json，macOS)格式相同 — 将 tvk_xxx 替换为您在 /settings 中获取的令牌。
将代码片段添加到仓库根目录的 .mcp.json 中，或使用 claude mcp add 以交互方式注册它。
通过环境变量覆盖 API 基础地址：
TRIGVALE_API_BASE_URL=https://api-dev.trigvale.com npx -y @trigvale/mcp默认基础地址为 https://api.trigvale.com（生产环境）。
“在搭建此项目之前，请使用用户的想法调用
validate_idea。如果结论是kill或pivot，请指出最薄弱的假设，并询问用户是否仍要继续。如果结论是test，请建议 Trigvale 在编写任何代码之前推荐的验证冲刺。”
构建成本已经变得很低。瓶颈不再是执行，而是决定执行什么。大多数想法在编写代码之前就应该被终止、转型或测试。validate_idea 使该决策成为您代理流程中一个明确、可重复的步骤，并提供您可以向用户展示的确定性结论和带有来源标记的证据。
计划 | 月费 | 包含内容 |
Free | $0 | 每月 1 次结论，精简预览 |
Starter | $12 | 每月 10 次结论，完整摘要 |
Validate | $39 | 实时证据管道 + 优化建议 + 冲刺生成器 |
Integrate | $99 | 以上所有内容 + 此 MCP 服务器的代理令牌 |
此 MCP 所需的代理令牌仅限 Integrate 计划。每次调用都会在服务器端重新检查权限（针对每个 /agent/v1/evaluate 调用）——降级会立即生效。
完整定价：https://trigvale.com/pricing
这是 Trigvale 私有代码仓库中 MCP 服务器源码的只读镜像（其余部分包含专有的评估准则、基础设施和计费逻辑）。版本发布在私有端进行并同步到此处。欢迎提交 PR —— 请通过 https://trigvale.com/contact 而不是 GitHub Issues 提交问题，以便我们将其纳入支持工作流。
MIT — sydacos GmbH, 2026.
Validate a startup idea before scaffolding code for it. Returns a scorecard (10-dimension Venture Readiness Score, 0–100), an evidence brief tagged by source (observed / inferred / AI / claim), a kill / pivot / test / build verdict, founder-fit calibration deltas if a skill graph is on file, and 1–3 archetype assignments with structural cautions for that cluster. Call this BEFORE writing project scaffolding when the user is greenfield-building from a raw idea.
| Name | Required | Description | Default |
|---|---|---|---|
| rawIdea | Yes | The idea to validate. Free-form text; mess is fine — Trigvale normalizes it. Be specific about who has the pain, who pays, and why now. | |
| save | No | When true, persist the brief to the user's vault so they can revisit it at trigvale.com/ideas/{id}. Default true. Set false for ephemeral checks. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description fully carries the burden of behavioral disclosure. It lists exact return values (scorecard, evidence brief, verdict, deltas, archetypes) and mentions conditional behavior ('if a skill graph is on file'). It also notes the persistence behavior tied to the `save` parameter.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is compact but includes necessary detail. It front-loads the core action ('Validate a startup idea before scaffolding code for it.'), then lists outputs efficiently. A very slight verbosity in listing all outputs prevents a perfect 5, but it remains well-structured.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's moderate complexity (2 parameters, no output schema, no annotations), the description provides sufficient context: it explains the full return structure, conditional behavior, and persistence options. No obvious gaps remain for an LLM to select and invoke the tool correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, and the description adds valuable usage hints beyond the schema: for `rawIdea` it advises specificity and normalizes messiness; for `save` it explains default behavior and ephemeral usage. This enriches the agent's understanding of each parameter.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description opens with a clear action-verb+resource: 'Validate a startup idea before scaffolding code for it.' It unambiguously states the tool's function and context, distinguishing it from potential siblings by specifying the prerequisite step.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly tells when to use the tool: 'Call this BEFORE writing project scaffolding when the user is greenfield-building from a raw idea.' While it doesn't mention when not to use it or alternatives, the guidance is clear and actionable given no sibling tools exist.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0validate_ideaScored across 1 tool
With only one tool, there is no possibility of confusion between tools. The single tool has a distinct and clear purpose.
The single tool 'validate_idea' follows a clear verb_noun pattern and is descriptive. However, with only one tool, consistency across multiple tools cannot be fully assessed.
One tool is borderline for the implied domain of startup building. While the validation step is specific, a broader lifecycle (e.g., scaffolding) is suggested but not covered, making the count feel thin.
The tool covers only idea validation. Obvious gaps exist, such as tools for scaffolding, iteration, or testing, which are referenced in the description but not provided. This will likely cause agent failures after validation.
Score, validate, and pressure-test startup ideas with AI from Claude or any MCP agent.
Rule-based product method: scores an idea, finds its riskiest assumption, sequences the build.
Validate startup ideas with your AI assistant through Startup Ignition ToolSuite.
Score business ideas on six live market signals. Launch Readiness Score (LRS) for AI agents.
