Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
作者による説明(英語の原文)
Read-only MCP gateway to 194 tokenized US equities on Robinhood Chain: quotes, holders, whale moves.
この3つを1つのスコアにまとめることはしません。意見、測定、人気はそれぞれ別のものだからです。
このサーバーのやさしい解説はまだ書いていません。以下はレジストリから読み取った事実です。
使っているAIアプリを選んで、設定をコピーしてください。
必要なもの
どのアプリを使っていますか?
上でアプリを選んでください。選んだアプリを覚えておくので、次からはボタンに「…にインストール」と表示されます。
Claude Desktop → Settings → Developer → Edit Config を開き、claude_desktop_config.json にこれを貼り付けます
{
"mcpServers": {
"arcus-agent-gateway": {
"command": "uvx",
"args": [
"arcus-agent-gateway"
]
}
}
}ターミナルでこれを実行します
claude mcp add arcus-agent-gateway -- uvx arcus-agent-gatewayプロジェクトの .cursor/mcp.json、または ~/.cursor/mcp.json に貼り付けます
{
"mcpServers": {
"arcus-agent-gateway": {
"command": "uvx",
"args": [
"arcus-agent-gateway"
]
}
}
}ブラウザがアプリを開いてよいか確認し、そのあとアプリが設定を表示して確認を求めます。キーの値は空欄のままなので、自分で入力してください。
プロジェクトの .vscode/mcp.json に貼り付けます
{
"servers": {
"arcus-agent-gateway": {
"type": "stdio",
"command": "uvx",
"args": [
"arcus-agent-gateway"
]
}
}
}ブラウザがアプリを開いてよいか確認し、そのあとアプリが設定を表示して確認を求めます。キーの値は空欄のままなので、自分で入力してください。
これはパソコンで動くサーバーですが、ChatGPTがつなげるのはインターネット上のサーバーだけです。ほかのアプリを選んでください。
選んだアプリは、この端末のCookieに保存されます。アカウントは不要です。
この設定は、作者のレジストリ登録情報から作成しています。READMEと違う場合は、READMEに従ってください。
パソコンで動くサーバーです。インストールすると、ツール一覧がAIアプリに表示されます。
誰でも見られる情報をチェックしており、コードは読んでいません。当てはまらない項目は数に入れていません。 · 2026年11月8日にチェック
レジストリ名は、リポジトリを所有するGitHubアカウント alekskram で確認済みです
このリポジトリはまだGitHubから読み取っていません
このリポジトリはまだGitHubから読み取っていません
これはパソコンで動くため、インターネット上のエンドポイントはありません
パソコンで動き、代わりにローカルの設定を使います
パソコンで動くサーバーです。ツール一覧を読み取れるのは、インターネット上のエンドポイントがある場合だけです
レジストリでこのカテゴリにある有効なすべてのサーバーと比較しています。スターは、公開者がリポジトリの所有者である場合だけ数えます。
当てはまる基本チェック1項目のうち1項目に合格しました。コードは読んでいません。キーやトークンを追加する前に、サーバーがどんなアクセスを求めているかを確認し、必要な権限だけを持つキーを使ってください。
レジストリには必須のキーは載っていません。ただし、最初に使うときにサインインを求めるサーバーもあります。
あなたのパソコンで動きます。必要なときにAIアプリが起動し、データは作者のサーバーを通りません(サーバー自体がオンラインサービスを呼び出す場合を除きます)。
GitHubリポジトリの最終更新は–、最新バージョン(0.2.4)がレジストリに届いたのは先月です。
Claude Desktop、Claude Code、Cursor、VS CodeなどのMCP対応アプリで使えます。パソコンにPython (uv)も必要です。
同じカテゴリには:Spec Driven Development, Remote Desktop Commander, Tool Guardian
同じカテゴリで、ほかの作者が作ったサーバー。スターの多い順。
カードの見方:★ GitHubスターi3つの柱:レビュー · チェック · 利用状況i更新=最新バージョンをレジストリに公開した日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.
Too many MCP servers eating your context? A router that loads each tool's schema only when needed.
作者のリポジトリから書かれたとおりにコピーしたものです。PopNx MCPは内容を確認・推奨していません。HTML、画像、スクリプトは取り除いています。リンクはこのサイトの外へ移動します。
Robinhood Chain (Arcus) carries 194 tokenized US equities on-chain, and its public API answers questions most people don't think to ask: who actually holds AAPL, what a wallet's stock portfolio looks like, which way the whales moved NVDA today. This MCP server hands all of that to an AI agent. Read-only, keyless, no writes; every tool is a GET against api.robinhood.com/rhj plus two public on-chain sources, cached and rate-limited so an enthusiastic agent can't hammer the upstream.
Ask it who holds AAPL and you get the top holder addresses with their share of total supply, flagged contract or wallet. One holder contract turns out to sit on 35% of NVDA; that is the kind of concentration risk a quote page will never show you.
wallet_holdings reads any address across all 194 equities and values the positions it has fresh quotes for. transfer_history walks recent ERC-20 Transfers with a min_value filter, which is how you catch the large prints instead of the dust.
Prices are the trap here, so they ship split-safe: raw and multiplier-adjusted values travel together on every quote, and a pending split (NVDA queues a 4.0 multiplier for 2026-11) comes with a warning and the effective time. A morning scan of market health, halted tokens and 13-sector averages costs two cheap calls.
Full walkthroughs with real outputs: examples/use-cases.md.
Claude Code:
claude mcp add arcus -- uvx arcus-agent-gateway
Run over stdio (the default, for local agents):
uvx arcus-agent-gateway
Standard config for Claude Desktop / Cursor (claude_desktop_config.json / .cursor/mcp.json):
{
"mcpServers": { "arcus": { "command": "uvx", "args": ["arcus-agent-gateway"] } } }
Codex (~/.codex/config.toml)
[mcp_servers.arcus]
command = "uvx" args = ["arcus-agent-gateway"]
</details>
ZCode : register the server and copy the agent skill
# 1) start the gateway (keep it running)
uvx arcus-agent-gateway --http --port 8902 &
python3 - <<'PY' import json, os p = os.path.expanduser("~/.zcode/cli/config.json") os.makedirs(os.path.dirname(p), exist_ok=True) cfg = json.load(open(p)) if os.path.exists(p) else {} cfg.setdefault("mcp", {}).setdefault("servers", {})["arcus"] = { "type": "http", "url": "http://127.0.0.1:8902/mcp"} json.dump(cfg, open(p, "w"), indent=2) print("arcus MCP server registered:", p) PY
git clone -q --depth 1 https://github.com/alekskram/arcus-agent-gateway /tmp/aag cp -r /tmp/aag/.agents/skills/arcus-gateway ~/.zcode/skills/ && rm -rf /tmp/aag echo "ZCode setup done, restart your session and call any arcus tool"
</details>
Hosted form, streamable HTTP on port 8902:
uvx arcus-agent-gateway --http # 127.0.0.1:8902
curl http://127.0.0.1:8902/health # -> {"ok": true, "service": "arcus-agent-gateway"}
All 13 tools are read-only (annotated readOnlyHint: true). Names and parameters are exactly as registered by arcus_mcp/server.py.
| # | Tool | Signature | What it does |
|---|---|---|---|
| 1 | token_list | token_list(status="ACTIVE", limit=100) | Tokenized equities, one row per token (symbol, name, status, multiplier, tradable); status filters the ASSET_STATUS_* prefix, 'ALL' disables. Start here for valid symbols. |
| 2 | quote | quote(symbol) | Live quote joined with asset metadata: raw + multiplier-adjusted bid/ask/spread, is_halted, trading capabilities, multiplier block. Unknown symbol raises with a pointer to token_list(). |
| 3 | quotes | quotes(symbols) | Batch of quote() rows, max 20 per call (more raises). Unknown symbols land in errors without failing the batch. Requests run in parallel (semaphore 8); 10 cold symbols take 0.6-1 s instead of ~3 s. |
| 4 | token_detail | token_detail(symbol) | Full dossier: contract/chain/ISIN metadata, embedded quote, last 5 corporate actions, multiplier block with history note, warnings (pending split). |
| 5 | market_status | market_status() | Market-wide health from assets only (never fetches 194 prices): totals, untradable count, cached-halted list, extended-hours estimate. |
| 6 | corporate_actions | corporate_actions(symbol=None, limit=10) | Splits/dividends across all tokens or for one symbol; tolerant to the API's field-name variants. |
| 7 | search | search(query, limit=10) | Local fuzzy search over the token list; apple → AAPL; top limit (cap 50) with scores and sectors. |
| 8 | sector_view | sector_view(warm=False, sector=None) | 13-sector static map with sizes and multiplier-adjusted sector averages. Default (warm=False): caches only, zero requests, warmed: false. warm=True, sector="...": fans out fresh quotes for that sector only and reports requests_made. |
| 9 | onchain_info | onchain_info(symbol) | On-chain footprint joined from three independent sources (each fails to a warnings[] entry, never silently): contract address, chain id (4663, Robinhood Chain), network, decimals, ISIN (REST) + total_supply via totalSupply() eth_call (source rpc) + holders_count, circulating_market_cap (source explorer). supply_crosscheck compares the REST-implied cap (total_supply × multiplier × mid) with the explorer's; >1% divergence → warning. Per-field source tags on every derived value. |
| 10 | price_history | price_history(symbol, timeframe="daily", limit=90) | OHLCV history from the optional recorder's local parquet store (see below). Honest degradation: missing pyarrow or data → actionable error dict, never a silent empty list. |
| 11 | holder_snapshot | holder_snapshot(symbol, limit=20) | Top holders of a token's contract from the Blockscout explorer (one page, max 50 rows, 600 s cache). Rows: address, value (float token units), share_pct = value / total_supply × 100, is_contract. total_supply from the RPC with an explorer fallback (source-tagged); no supply at all → share_pct: null + warning. Errors → error dict with kind + hint. |
| 12 | wallet_holdings | wallet_holdings(address) | Which of the 194 tokenized equities a wallet holds (explorer token-balances ∩ assets() universe). Rows: symbol, name, value (float token units). est_position_usd / portfolio_usd_total computed ONLY from quotes already in the price cache (no fan-out); missing/stale quotes → null estimates + explanatory note. Cached 120 s. |
| 13 | transfer_history | transfer_history(symbol, limit=25, min_value=None) | Recent ERC-20 Transfer events from the public RPC's adaptive walk-back (windows start 48 blocks wide, shrink 48→32→16→8 on archive 403s, ≤14 getLogs requests — (see On-chain sources & limits)). Rows (newest first): ts (ISO, from the log's own blockTimestamp), from, to, value (float), tx_hash, block. min_value filters in token units; window exhausted with 0 logs → explicit note pointing at the explorer. Cached 60 s. |
| — | *watchlist* | — | Not a tool. Price tracking is done by your agent's scheduler (cron) calling quotes() on an interval — (see .agents/skills/arcus-gateway/SKILL.md). |
Robinhood Chain tokens carry a multiplier, the corporate-action adjustment factor for the token contract (1.0 = untouched). Splits change it; for example NVDA's 2026-11 split queues pendingMultiplier: "4.0".
bid/ask from /prices/{symbol} are in token-contract units and are not multiplier-adjusted.price_adjusted = round(price_raw × currentMultiplier, 6).bid_raw/ask_raw/spread_raw *and* bid_adjusted/ask_adjusted/ mid_adjusted next to the multiplier block, never one without the other.*_adjusted fields.Worked example (live fixture, 2026-09-03):
AAPL currentMultiplier = 1.000566080061092436
bid_raw = 327.77 → bid_adjusted = round(327.77 × 1.000566…, 6) = 327.955544 ask_raw = 327.78 → ask_adjusted = 327.965550 mid mid_adjusted = 327.960547
Pending split warning. When pendingMultiplier is queued (non-empty) and differs from the current one, token_detail() adds a warning like pending split: 1→4.0 on 2026-11-06T00:00:00Z, and quote()'s multiplier block exposes pending + effective_time. After the split lands, raw prices jump by the ratio while *_adjusted fields stay comparable, another reason to always read adjusted values next to the multiplier.
api.robinhood.com/rhj + the public RPC are open, and every agent hitting them directly rediscovers the same traps:
| Raw sources give you | You would have to build |
|---|---|
RAW, non-multiplier-adjusted prices (bid/ask in contract units) | the multiplier math, raw/adjusted pairs on every quote, pending-split detection with effective times |
| no price history endpoint at all | a recorder (opt-in here): 5-min snapshots → parquet → idempotent daily OHLCV rollup |
| 60 req/s upstream limit | a polite rate-limited client (≤50 req/s), per-endpoint caches, parallel batched quotes |
| RPC archive window that 403s outside ~45–60 blocks behind head | adaptive walk-back (48→32→16→8 block windows, ≤14 getLogs) |
| a Blockscout explorer behind a Cloudflare UA check, 40 s hangs on contract wallets | browser UA, timeouts, honest error/warnings[] degradation, never a silent empty list |
429/502/503/504, network errors) are retried up to 3 times with 2s × (attempt+1) backoff./assets 5 min, /prices/{symbol} 15 s, /corporate-actions 1 h. market_status() and sector_view() are computed from caches and assets only; they never fan out 194 price requests.The v0.2 on-chain tools read two keyless public sources next to the REST API. Both are free, rate-limited and partially restricted. Every tool above degrades honestly (per-field omission + warnings[] / error dicts), never with a silent empty answer.
robinhood-rpc.publicnode.com, override with ARCUS_RPC_URL): eth_call (e.g. totalSupply()) works normally. **eth_getLogs only answers inside a floating ~45–60-block window behin冒頭だけを表示しています。続きはリポジトリで読めます