Scan the tape
Sentiment cycle, leading theme, who's driving it — institutions or hot money. One question, a grounded read of the whole session.
Your AI already knows how to reason. What it lacks is the tape. One remote endpoint plugs EchoPai into Claude Code, Cursor, Workbuddy — any AI that speaks MCP — and hands it 40+ live A-share tools it calls mid-conversation. Nothing to install. Every answer traces back to a tool call.
Not a chatbot with opinions — an agent with data. Every claim traces back to a tool call.
Sentiment cycle, leading theme, who's driving it — institutions or hot money. One question, a grounded read of the whole session.
Why it moved, who wrote about it, what the filings say. Quotes, news, broker views and announcements, joined in seconds.
Full-text sell-side reports and analyst views, searchable and attributable — your agent reads the 40 pages so you read four lines.
Pool, streaks, promotion rate, attribution — yesterday's board reconstructed with numbers, not narratives.
300+ concept boards with constituents, flows, and news — watch a theme form, crowd, and roll over before the index tells you.
Announcements, holder trades, lockup expiries — searchable the moment they drop, with the ownership context that says whether they matter.
The same production APIs behind the EchoPai terminal, served to your AI over MCP.
41 tools today — the catalog grows with the data platform. Full schemas are served over MCP; your AI discovers them on its own.
Point your AI at the remote endpoint. No SDK, no local process, no upgrades to babysit — the tool catalog is served live.
claude mcp add --transport http echopai https://mcp.huishengpai.com/mcp. Cursor, VS Code and Gemini CLI take the same URL as one JSON block.Add the remote endpoint, ask your first question — the whole setup takes about twenty seconds, and installs nothing.
Prefer a visual workspace? Explore the EchoPai terminal →