
echopai-mcp
io.github.echopai/echopai-mcp
China equity & quantitative research MCP server with 50 tools for reports, filings & sentiment.
Documentation
EchoPai MCP Server · 回声投研
Turn any AI into an event-driven equity research analyst.
Equip Claude, Cursor, VS Code, Gemini CLI, or any MCP-compatible agent with 50 real-time China equity (A-share) & quantitative research tools — full-text broker reports, individual analyst attribution, corporate filings, quantified sentiment cycles, and limit-up analytics, with traceable citations on every answer.
This repository provides a complete Node.js MCP Server built on @modelcontextprotocol/sdk and zod (src/index.js, src/tools.js, src/client.js) wrapping EchoPai's Open Platform REST APIs, alongside a hosted Streamable HTTP endpoint (https://mcp.huishengpai.com/mcp) for zero-install OAuth setup.
Official Portal Matrix
| Resource | Global (EN) | China Mainland (中文 · 回声投研) |
|---|---|---|
| Official Website | https://echopai.com | https://www.huishengpai.com |
| Web Research Terminal | https://app.echopai.com | https://huishengpai.com/go/app |
| API Keys & OAuth Account | https://echopai.com/go/app/agent | https://huishengpai.com/go/app/agent |
| Remote MCP Endpoint | https://mcp.huishengpai.com/mcp | https://mcp.huishengpai.com/mcp |
| Automated AI Setup Prompt | https://echopai.com/agent-setup/prompt.md | https://www.huishengpai.com/agent-setup/prompt.md |
Why EchoPai MCP?
Most open-source stock MCP servers only wrap basic price quotes or K-line endpoints. EchoPai connects your AI to institutional-grade event-driven research infrastructure:
| Capability | Basic Price / Free Scripts | EchoPai MCP (回声投研) |
|---|---|---|
| Broker Research | None or unindexed PDFs | Full-text broker reports (report) + analyst-level attribution weighted by historical accuracy (views) |
| Market Events & Filings | Delayed news headlines | Real-time announcements (announcements_*), forward-looking event calendar, and share lockup schedules (ownership_lockups) |
| Market Microstructure | Raw OHLCV only | Limit-up pool & attribution (limit_up_*), quantified sentiment cycles (sentiment_*), and 300+ concept boards (concepts_*) |
| Deployment & Auth | Local Python/Node setup & fragile cookies | Zero-install remote Streamable HTTP with Browser OAuth 2.0 or API Key; tools stay updated server-side |
What It Looks Like in Action
Once connected, simply ask complex research questions in natural language. Your AI chains multiple EchoPai tools automatically and cites every claim:
Example 1: Morning Broker Research Synthesis ("Read all compute research notes from this morning and summarize")
AI automatically calls:
views→report→digest→search
- Consensus (3 top institutes aligned): Domestic training chip shipments revised up (40% → 55% YoY); 1.6T optical module ramp pulled forward to Q4.
- Divergence: One sell-side team flags cautious CapEx guidance and yield/delivery cadence differences in Q2.
- Analyst Weighting: Automatically surfaces each analyst's 90-day track record and directional hit rate so you know whose call carries more weight.
Example 2: Intraday Surge Attribution ("Why is Cambricon surging today?")
AI automatically calls:
quote→news→views→announcements→sentiment
- Market Microstructure: +9.98% on 12.8B CNY turnover (8.2% turnover rate, 98th percentile over 60 days; volume z-score +3.1; main net inflow +1.1B CNY).
- Catalyst: Equity incentive draft announced last night (grant price ~66% of market, covering 85% of staff, 2026–2028 revenue & profit targets).
- Broker Follow-up: 3 morning notes raised earnings forecasts / target prices.
- Risk Check: Verified zero reduction announcements within 30 days, zero lockup expirations within 45 days, and no regulatory inquiry letters.
Quick Start (20 Seconds)
Remote Endpoint (Streamable HTTP):
https://mcp.huishengpai.com/mcp
Supports both Zero-Key Browser OAuth 2.0 and Bearer API Key (eps_live_...).
Claude Code
Option 1: Browser OAuth (no API key needed)
claude mcp add --transport http echopai https://mcp.huishengpai.com/mcp
Select echopai in /mcp to complete browser authorization.
Option 2: API Key
claude mcp add --transport http echopai https://mcp.huishengpai.com/mcp --header "Authorization: Bearer <YOUR_KEY>"
Cursor
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"echopai": {
"url": "https://mcp.huishengpai.com/mcp",
"headers": {
"Authorization": "Bearer <YOUR_KEY>"
}
}
}
}
VS Code (GitHub Copilot)
Add to .vscode/mcp.json:
{
"servers": {
"echopai": {
"type": "http",
"url": "https://mcp.huishengpai.com/mcp",
"headers": {
"Authorization": "Bearer <YOUR_KEY>"
}
}
}
}
Claude Desktop / claude.ai / WorkBuddy
Open Settings → Connectors, add https://mcp.huishengpai.com/mcp, and click Authorize via browser login.
Gemini CLI
gemini mcp add --transport http --header "Authorization: Bearer <YOUR_KEY>" echopai https://mcp.huishengpai.com/mcp
See Client Configuration Guide for project-shared configs, environment variable tips, and headless CI/Cron usage.
Tool Directory (50 Tools)
All 50 tools are explicitly registered with zod input schemas in src/tools.js and call EchoPai's Open Platform /v1 REST APIs via src/client.js (with automatic failover between api.huishengpai.com and api.echopai.com).
| Category | Tools |
|---|---|
Account & Market Meta (×4) | whoami, usage, market_status, trading_days |
Research & Views (×3) | views, report, digest |
News, Announcements & Events (×7) | news, announcements_feed, announcements_search, announcements_stock, announcements_detail, events_upcoming, events_timeline |
Financials & Earnings (×6) | financials_reports, financials_series, financials_pit, financials_quote_snapshot, earnings_show, earnings_recent |
Ownership & Lockups (×2) | ownership_show, ownership_lockups |
Market Quotes, Technicals & Scan (×12) | quote, bars_batch, chart, index_snapshot, market_movers, scan, hot, search, lookup, technicals, fundflow, dragon_tiger_daily |
Limit-up Analytics (×4) | limit_up_pool, limit_up_analysis, limit_up_history, limit_up_summary |
Market Sentiment (×4) | sentiment_overview, sentiment_breadth, sentiment_turnover, sentiment_pct_distribution |
Concept Boards (×8) | concepts_list, concepts_show, concepts_snapshot, concepts_news, concepts_views, concepts_alerts, concepts_minute_bars, concepts_daily_bars |
For detailed descriptions and usage rules, see Tool Reference.
Local stdio MCP Server & CLI
Run this repository directly as a standalone local stdio MCP server powered by @modelcontextprotocol/sdk and zod (src/index.js, requires Node.js ≥ 20):
{
"mcpServers": {
"echopai": {
"command": "npx",
"args": ["-y", "@echopai/mcp"],
"env": {
"ECHOPAI_API_KEY": "eps_live_YOUR_KEY_HERE"
}
}
}
}
Or run directly from a local clone:
node src/index.js --key eps_live_YOUR_KEY_HERE
For shell pipelines (echopai ... | jq) or cron jobs, you can also install the full echopai CLI package from npm:
npm install -g echopai
echopai --version
Automated AI Setup
Paste either official setup prompt URL directly to your AI assistant and let it configure everything automatically:
- Global: https://echopai.com/agent-setup/prompt.md
- China Mainland: https://www.huishengpai.com/agent-setup/prompt.md
About EchoPai & Disclaimer
Founded in 2022 with a distributed team across New York, Singapore, and Asia-Pacific, EchoPai (回声投研) builds event-driven quantitative research infrastructure based on the thesis that the market is the echo of events.
- Global Website: https://echopai.com
- China Website: https://www.huishengpai.com
- Web Terminal: https://app.echopai.com
Disclaimer: EchoPai is a financial data and research information tool—not a registered investment advisor or broker-dealer. Nothing provided by the MCP server, API, or documentation constitutes investment advice or solicitation to buy or sell any securities.
License
MIT © 2026 EchoPai LLC
@echopai/mcpnpm install @echopai/mcp