Verity MCP Server
Enables AI agents to invoke six pre-trade integrity check skills: detect coordinated activity, score market subjects, cross-check claims, and flag disinformation patterns.
README
Verity MCP Server
Public MCP (Model Context Protocol) server for Verity skills. Six callable endpoints your AI agent can invoke before any trade: detect coordinated activity, score market subjects, cross-check claims, flag disinformation patterns.
Production endpoint: https://mcp.verityskills.com/mcp
Why Verity skills
ApeWisdom has a dashboard. StockTwits has a feed. Verity has skills. The pre-trade integrity check most AI workflows are missing. Connect once; every Verity capability becomes a function call from inside your Claude, Cursor, LangChain, or LlamaIndex workflow.
Every skill needs a Verity API key. Issue one in seconds at https://verityskills.com/account/api-keys.
Install
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows). Native HTTP transport (Claude Desktop 2026 builds):
{
"mcpServers": {
"verity": {
"type": "http",
"url": "https://mcp.verityskills.com/mcp",
"headers": {
"Authorization": "Bearer vtk_<your-token>"
}
}
}
}
If your Claude Desktop is older and rejects type: "http", use the mcp-remote stdio bridge. See docs/CLAUDE_DESKTOP.md for the full guide. Fully quit Claude Desktop (Cmd-Q on macOS) before relaunch.
Cursor
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"verity": {
"url": "https://mcp.verityskills.com/mcp",
"headers": {
"Authorization": "Bearer vtk_<your-token>"
}
}
}
}
Restart Cursor. See docs/CURSOR.md for the full guide.
Direct API (curl, no MCP client required)
curl -sS -X POST https://mcp.verityskills.com/mcp \
-H 'content-type: application/json' \
-H 'Authorization: Bearer vtk_your_key_here' \
-d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "coordination-heat",
"arguments": { "subject": "GME" }
}
}'
All six skills require an API key issued at https://verityskills.com/account/api-keys.
Other agent runtimes
Any MCP-compliant agent runtime works. The endpoint is https://mcp.verityskills.com/mcp over HTTP+SSE (the SSE channel is a future addition; HTTP POST works today). Add Authorization: Bearer vtk_<token> for every skill.
Available skills
| Skill | Auth | Capability |
|---|---|---|
coordination-heat |
Required | Score a topic, ticker, or narrative for coordinated activity. |
verity-score |
Required | Composite integrity score for a market subject. |
morning-brief |
Required | Daily briefing of overnight signals across a watchlist. |
verity-scan |
Required | Real-time scan of a topic, ticker, or claim. |
cross-check-alert |
Required | Cross-reference a claim against monitored sources. |
disinfo-alert |
Required | Monitor a topic or ticker for disinformation patterns. |
Marketplace listings
- Smithery.ai (pending review)
- Cursor Directory (pending review)
- PulseMCP, Glama, MCP.so (pending review)
Architecture
Thin bearer-token proxy on Cloudflare Workers. Zero state, zero DB credentials.
- Translates
Authorization: Bearer vtk_*from MCP clients into the upstreamx-verity-keyheader for verityskills.com skill routes. - JSON-RPC 2.0 framing per the MCP spec (
2025-03-26). - 402
TRIAL_CAP_REACHEDresponses are rewritten to inject an absoluteupgrade_urlso MCP clients can route trial-cap users to the right upgrade page. - 5xx upstream responses are wrapped as JSON-RPC -32603 envelopes with HTTP 200 to keep transport-layer retry semantics correct.
- Per-tool kill switch via the
MCP_TOOLS_DISABLEDWrangler secret. Global kill switch viaMCP_KILL_SWITCH=on. - Sentry instrumentation via @sentry/cloudflare. Bearer tokens are scrubbed from any captured event via a
beforeSendhook.
Operations
| Concern | Source |
|---|---|
| Deploy + secrets + DNS | docs/DEPLOY_RUNBOOK.md |
| Post-deploy smoke | scripts/post-deploy-smoke.sh <base-url> |
| Synthetic monitoring | 5-minute cron at cchurctrip/verity:app/api/cron/mcp-synthetic-smoke/route.ts |
| Health endpoint | GET https://mcp.verityskills.com/health |
| Marketplace submission | docs/SMITHERY.md, docs/CURSOR.md |
Development
git clone https://github.com/cchurctrip/verity-mcp
cd verity-mcp
npm install
npm test # 237 unit + property + integration tests
npm run lint
npm run typecheck
npm run check-manifest
npx wrangler dev # local worker on http://localhost:8787
License
MIT. See LICENSE.
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。