Proposal Generator MCP
Generates professional client proposals including timeline, tech stack, price estimate, and milestone breakdown from client requirements.
README
Proposal Generator MCP
An MCP (Model Context Protocol) server that turns client requirements into a timeline, tech stack, price estimate, milestone breakdown, and a ready-to-send Markdown proposal document — built for freelancers and agencies.
{
"clientName": "Green Valley Fertilizers",
"projectType": "ecommerce",
"features": ["auth", "ecommerce-catalog", "payments", "admin-dashboard", "search"],
"region": "pakistan"
}
Use it from Claude Desktop, Claude Code, Cursor, or any MCP-compatible client.
Tools
| Tool | Returns |
|---|---|
generate_proposal |
Full pipeline: JSON summary + Markdown proposal document |
estimate_price_and_timeline |
Quick estimate: hours, price range, timeline |
recommend_tech_stack |
Tech stack suggestion based on project type + features |
list_feature_catalog |
Every recognized feature key with its hour estimate and category |
Install & run
Option A — from source (local development)
git clone <repo-url> proposal-generator
cd proposal-generator
bun install
bun run build
Start the server:
bun start
Option B — from npm (published package)
npm install -g @proposal-generator/server
# or
bun install -g @proposal-generator/server
Run it:
proposal-generator-mcp
Scripts reference
| Script | Description |
|---|---|
bun run build |
Build with tsdown (ESM + .d.ts + sourcemap) |
bun run dev |
Watch mode rebuild on changes |
bun run lint |
Auto-fix lint & formatting issues (biome) |
bun run lint:check |
Check lint without writing |
bun run typecheck |
tsc --noEmit type verification |
bun run release |
Build, bump version, publish to npm |
Configure your MCP client
Claude Desktop
Edit claude_desktop_config.json (location varies by OS):
Local source install:
{
"mcpServers": {
"proposal-generator": {
"command": "bun",
"args": ["run", "/absolute/path/to/proposal-generator/packages/server/dist/index.mjs"]
}
}
}
Global npm install:
{
"mcpServers": {
"proposal-generator": {
"command": "npx",
"args": ["@proposal-generator/server"]
}
}
}
Claude Code
# Local source install
claude mcp add proposal-generator -- bun run /absolute/path/to/proposal-generator/packages/server/dist/index.mjs
# Global npm install
claude mcp add proposal-generator -- npx @proposal-generator/server
Cursor
Settings → Features → MCP Servers → Add new MCP server:
| Field | Local | npm |
|---|---|---|
| Name | proposal-generator |
proposal-generator |
| Type | command |
command |
| Command | bun run /absolute/path/.../dist/index.mjs |
npx @proposal-generator/server |
opencode
If using opencode, add to your project's opencode.json:
{
"mcpServers": {
"proposal-generator": {
"command": "bun",
"args": ["run", "packages/server/dist/index.mjs"]
}
}
}
Any MCP client (stdio)
{
"mcpServers": {
"proposal-generator": {
"command": "bun",
"args": ["run", "/path/to/packages/server/dist/index.mjs"]
}
}
}
Customizing estimates
Everything that drives the numbers lives in two files — no logic changes needed:
| File | What to tune |
|---|---|
packages/server/src/data/featureCatalog.ts |
Feature keys, hour estimates, tech pulled in |
packages/server/src/data/rateCards.ts |
Regional hourly rates, base hours per project type, complexity multipliers |
The milestone template (phase names, payment splits) lives in packages/server/src/logic/estimator.ts.
The proposal document layout lives in packages/server/src/logic/proposalBuilder.ts.
After editing any of these, rebuild with bun run build.
Project structure
proposal-generator/
├── package.json # workspace root (bun monorepo)
├── biome.json # lint & format config
├── tsconfig.base.json # shared TypeScript config
├── opencode.json # opencode MCP config
├── .gitignore
├── README.md
│
└── packages/
└── server/ # @proposal-generator/server
├── package.json
├── tsconfig.json # extends tsconfig.base.json
├── tsdown.config.ts # build config
├── dist/ # build output (gitignored)
└── src/
├── index.ts # MCP server + tool definitions
├── data/
│ ├── featureCatalog.ts # feature -> hours/tech mapping
│ └── rateCards.ts # regional rates, base hours, multipliers
└── logic/
├── estimator.ts # core math: hours, price, timeline, milestones
└── proposalBuilder.ts # renders the Markdown proposal document
Publishing to npm
# Login (one-time)
npm login
# Bump version, tag, and publish
bun run release
This runs: bun run build → bumpp (version bump + git tag) → npm publish.
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。