Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
作者提供的说明(英文原文)
Live Base staking/lending yields (APY, TVL, vaults) as MCP tools; x402 pay-per-request tier
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
STAKECHAIN_API你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
适用于网页版、桌面版和手机版 Claude:
https://stakechain-api.vercel.app/mcp打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"stakechain": {
"command": "npx",
"args": [
"-y",
"stakechain-mcp"
],
"env": {
"STAKECHAIN_API": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add --transport http stakechain https://stakechain-api.vercel.app/mcpclaude mcp add stakechain --env STAKECHAIN_API=your-value -- npx -y stakechain-mcp运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"stakechain": {
"url": "https://stakechain-api.vercel.app/mcp"
}
}
}粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"stakechain": {
"command": "npx",
"args": [
"-y",
"stakechain-mcp"
],
"env": {
"STAKECHAIN_API": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"stakechain": {
"type": "http",
"url": "https://stakechain-api.vercel.app/mcp"
}
}
}粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"stakechain": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"stakechain-mcp"
],
"env": {
"STAKECHAIN_API": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
ChatGPT 只能连接互联网上的服务器:
https://stakechain-api.vercel.app/mcp菜单名称可能因套餐和版本不同而略有差异。
你的选择通过 Cookie 保存在这台设备上,无需账户。
STAKECHAIN_API env这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
工具名称和描述与服务器在 2026年11月6日 报告的完全一致。这些内容由作者撰写,不是我们写的;我们以纯文本形式显示。
list_base_yieldsSurvey staking and lending opportunities on Base: APY, TVL, project and asset for each. Use this for breadth — browsing the market, answering 'what is available', or filtering a wide set. It covers pools this server cannot deposit into as well as ones it can. Use something else w
best_yieldThe single highest-APY pool on Base for one asset, with junk-farm filters applied (minimum TVL, APY ceiling, no impermanent-loss exposure). Use this for a direct 'where is the best rate for X' question when one answer is wanted rather than a list. Use something else when: you wan
list_lending_vaultsThe Base lending shelf an agent can act on: Morpho ERC-4626 vaults, Aave V3 reserves, Moonwell markets and this project's own staking pools, each with the contract address to deposit into (`depositTo`), the asset's address and decimals, and a `risk` block (fee, age, concentration
market_profileRisk profile for one market, for the moment before capital is committed. Use this when a specific market has been chosen and the question is whether to go ahead. It is the one tool that answers 'what else should I know besides the rate'. Use something else when: you are still cho
yield_historyDaily APY and TVL history for one market, with min, median, max and volatility over the window. Use this for one question only: has this rate held, or did it appear yesterday? Call it before acting on any unusually high rate. Use something else when: you want the other risk facto
build_depositTurns a chosen market into signed-ready calldata: an ordered list of transactions to deposit into it or leave it, across Morpho vaults, Aave reserves, Moonwell markets and this project's staking pools — three different protocol interfaces behind one call. The approve step is incl
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月7日
注册名称已通过 GitHub 账户 Abstract36 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
在 182 毫秒内响应(HTTP 200)
无需登录
已读取 6 个工具描述,未发现针对 AI 的隐藏指令
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 3 项适用的基础检查中通过了 3 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
两种方式都可以:安装在你的电脑上,或者连接云端版本。
GitHub 仓库最后更新于–,最新版本(0.3.0)于7天前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,所以接受远程 MCP 服务器的应用也可以使用,例如开启开发者模式的 ChatGPT。
同一分类中有:Spec Driven Development, Remote Desktop Commander, Python MCP
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Local-only Python MCP server for the Files.com API.
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
## ⚠️ Retired 2026-11-05 — not maintained
This project is finished. The daily and weekly schedules are off, nothing is
being updated, and no claim below should be read in the present tense.
It was built to sell DeFi rate data to machines over x402. It worked, and it
earned $0.008 a month. Sweeping the entire x402 discovery index — 3,315
sellers — showed total market revenue of $96,010 per 30 days, of which
71.9% is one gift-card checkout endpoint, and 2,736 sellers earn under $1 a
month. There is no price at which selling data on that rail pays.
Read [POSTMORTEM.md](POSTMORTEM.md) for the numbers, the three
conclusions the data overturned, and what is worth taking from the code.
The deployed links below may still respond; they are not being watched.
Live: landing · dashboard · API · contract on BaseScan · Blockscout
STACH is tradeable: Uniswap v4 pool STACH/ETH live on Base (initial price 4,000,000 STACH/ETH) — routable through aggregators (LI.FI verified).
An open staking-yield experiment on Base. Live aggregator of staking and lending rates across Base protocols — readable by humans on a dashboard, consumable by AI agents over MCP, payable by machines over x402.
No seed round. No token presale. Just a running, open-source experiment. Concept and plan: docs/PIVOT-2026.md.
| Piece | What it does |
|---|---|
| api/ | Base yields aggregator with junk-farm filters, rate history and market profiles. Aave and Moonwell are read from their own contracts, not a third-party snapshot; x402 machine-payment endpoints |
| agents/ | MCP server (5 tools, on npm) + runnable recipes + a resident analyst that pays the public API for its own data |
| dashboard/ | React dashboard: Base Yields, Earn (Morpho, Aave, Moonwell), Borrow, Swap, STACH staking, Portfolio |
| contracts/ | STACH token + flexible/locked staking — audited & hardened (audit), plus the treasury timelock — 38 tests |
| website/ | Landing page (static, GitHub-Pages-ready) |
Legacy 2021–2023 material (Laravel site, old static site, branding, original docs) is kept under website-laravel/, website-static/, branding/, docs/.
Human endpoints (/yields, /vaults) are free. Machine endpoints (/v1/*) speak the x402 protocol with enforcement on: HTTP 402 + payment requirements → the client signs an EIP-3009 USDC authorization ($0.001/request) → it settles on-chain through the Coinbase facilitator, with our own EIP-3009 settler as the fallback (no accounts, no API keys either way). Revenue lands at the admin wallet; api/pay-client.js is a reference paying client.
Embeddable badges: live rates as SVG for any README — see docs/BADGES.md. No key, no signup.
API reference: docs/API.md — endpoints, the x402 flow, MCP tools and what each one answers.
Rate history: /history?pool= &days=30 answers the question a live APY cannot — has this rate held, or did it appear yesterday? Daily snapshots are committed to data/history/ in this repo, so the series is verifiable and survives us. Agents get it as the yield_history tool.
Pricing: ~$0.001/request (pure x402; the three data routes differ by a millionth of a dollar so the settled transfer names which one was bought), or subscriptions bought with a single x402 payment — Developer $19/30d, Professional $99/30d (GET /v1/subscribe/ returns a signed X-API-KEY that skips per-request payments). Public stats: /stats reports payment counts and revenue computed from on-chain USDC transfers — verifiable by anyone on BaseScan.
Real yield: collected USDC is distributed to STACH stakers through RevenueDistributor (verified; merkle epochs reproducible from public state via contracts/scripts/revenue-snapshot.js, 180-day sweep-back).
The yields aggregator is an MCP server, published on npm as stakechain-mcp. Add it to any MCP-compatible client:
{
"mcpServers": { "stakechain": { "command": "npx", "args": ["-y", "stakechain-mcp"] } } }
# 1. Yields API
cd api && npm install && npm start # :4000
cd dashboard && npm install --legacy-peer-deps npm start # :3000, reads the API
claude mcp add stakechain -- npx -y stakechain-mcp
cd contracts && npm install npx hardhat test # 38 tests npm run deploy:baseSepolia # free testnet
Deployed at the same address on both networks (verified via Sourcify, exact match):
0x82c5f9901CD9AEE3177053482b79aEA235507166 · Blockscout0x82c5…7166 — real STACH stays locked on Base (bridge docs)0x82c5f9901CD9AEE3177053482b79aEA235507166Initial holder / admin: 0x2Ab5A19aabc599C1a24CC0541ECB757b56976661. The dashboard targets mainnet by default; set REACT_APP_CHAIN=baseSepolia for the testnet.
Fixed supply 42M. At deployment: 3M held by the contract as the staking reward reserve (1M flexible / 2M locked over 360 days), 39M to the admin wallet. The admin wallet now holds none of it: on 2026-10-31 it moved 24.43M (58.16% of supply) into a Dutch auction and a Uniswap v4 pool initializer — see the table below and docs/TOKENOMICS.md. The 2022 contract had two critical bugs (recipient-burning transferFrom, unbacked reward minting) — fixed and regression-tested in 2026: docs/AUDIT-2026-10.md.
The distributor above is deployed and verified, but carries no meaningful revenue yet — /stats reports the real figure, including how much came from our own agent. Revenue first, distribution second: the mechanism exists so it can be checked, not because there is something to distribute.
Full detail, including lock schedules and the usage-rewards budget: docs/TOKENOMICS.md.
Every STACH is accounted for (snapshot 2026-11-02, verify on-chain):
| Holder | Address | STACH | Share |
|---|---|---|---|
| Dutch auction — 58.16% of supply was moved here and to the initializer below on 2026-10-31; terms not yet published, nothing sold so far | 0xBd0c…FecC7 | 12.21M | 29.08% |
| Uniswap v4 pool initializer for that auction | 0xf101…6000 | 12.21M | 29.08% |
| Uniswap v4 STACH/ETH pool liquidity (not locked) | 0x4985…2b2b | 8.14M | 19.39% |
| Timelocked treasury — locked until 2027-04-28, linear to 2028-10-19, no owner and no recovery path | 0x6425…D69C | 6.30M | 15.00% |
| Staking reward reserve (3M) + user stakes, held by the token contract | 0x82c5…7166 | 3.03M | 7.22% |
| Deployer (deploys contracts, pays usage rewards) | 0x2974…734F | ~100k | 0.24% |
| Treasury / operations — now empty | 0x2Ab5…6661 | 0 | 0% |
15% sits in a verified timelock with no owner, no pause and no recovery function, and 7.22% is the staking reward reserve the token contract can only pay to stakers. About 80% of supply has never been in public hands — the auction and its initializer hold 58.16% between them, and the auction's terms are not published yet, so we cannot tell you here when or at what price those tokens reach the market. Concentration flags remain entirely fair; this table exists so the claim can be checked rather than trusted.
This is an experiment, not "the future of finance". Everything here is MIT-licensed, the audit is public, the data endpoints filter out obvious degen traps (min TVL, APY ceiling, no-IL default). Nothing in this repository is financial advice.