Convex MCP Server
Plug live Convex macro data into Claude Desktop, Claude Code, Cursor, and any MCP-compatible agent. Six tools, no API key, anonymous access up to 100 requests per day.
Quick install
claude mcp add convextrade -- npx -y @convextrade/mcp-server
Edit claude_desktop_config.json ( ~/Library/Application Support/Claude/ on macOS, %APPDATA%\\Claude\\ on Windows):
{
"mcpServers": {
"convextrade": {
"command": "npx",
"args": ["-y", "@convextrade/mcp-server"]
}
}
}Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"convextrade": {
"command": "npx",
"args": ["-y", "@convextrade/mcp-server"]
}
}
}Available tools
| Tool | Args | Description |
|---|---|---|
| get_metric | ticker | Current value, 7-day and 30-day change for any of ~150 Convex-tracked tickers, from SPY and VIX to T10Y2Y, HY OAS, and WALCL. |
| get_regime | (none) | Convex macro regime classification: overall regime, trajectory, transition target, scenario probabilities. |
| get_crpi | (none) | Convex Recession Probability Index value and channel-level breakdown (yield curve, Sahm rule, claims, HY credit, LEI). |
| get_macro_summary | (none) | Full dashboard snapshot: CVRP, CNLI, CRAI, NVI, key rates, credit spreads, volatility, narrative state. |
| list_scenarios | status?, limit?, tag? | Scenarios tracked by the Convex research desk with thesis, status, asset exposure, and base-rate probability. |
| search_glossary | query, limit? | Search the Convex glossary for authoritative macro definitions. Matches against term, aliases, and summary. |
| get_latest_articles | limit? | Latest Convex research-desk articles with title, topic, published timestamp, and summary. Ideal for syndication or "what is Convex writing about". |
| get_signals | (none) | Composite signal summary: CVRP, CNLI, CRAI, NVI plus key market signals. Quick read on aggregate risk and liquidity. |
| get_narrative_velocity | topN? | Current NVI value plus top trending macro narrative terms across 46 curated RSS feeds. Detects narrative regime shifts before price action. |
| ask_convex | question | Natural-language macro or markets question, answered by Convex AI with live data from 200+ indicators, current regime, and scenario ledger. Best for open-ended prompts. |
Example conversation
What does Convex think recession probability looks like right now, and which channels are driving it?
Convex CVRP is at 46 (Elevated band). The heaviest-weighted channel is the HY credit spread z-score at 12 of 20 points, followed by initial claims momentum at 11. The yield-curve channel has normalised to 4. The Sahm rule is at 7 of 20, and the LEI channel is at 12. Methodology and channel-level data: convextrade.com/methodology/indicators/cvrp.
FAQ
What is the Model Context Protocol?
MCP is an open standard introduced by Anthropic in late 2024 for connecting LLMs to external tools and data. An MCP server exposes structured tools that an agent can call; the agent decides when to call which tool based on the user’s question.
What does the Convex MCP server give me?
Ten tools that return live, structured, cited Convex macro data: individual metrics, the full dashboard snapshot, CVRP with channel breakdown, regime classification, tracked scenarios, glossary search, latest research articles, composite signal summary, narrative-velocity reading, and a natural-language ask endpoint. Each response includes a link back to the underlying methodology page.
Does it require an API key?
No. The server calls the public Convex HTTPS API anonymously. Rate-limited to 100 requests per day per IP. For higher limits, contact us.
What LLMs work with it?
Any MCP-compatible client: Claude Desktop, Claude Code, Cursor, Continue, Zed, or custom agents built on the MCP SDK. The protocol is transport-agnostic; this server uses stdio, which is supported by every mainstream client.
What is the source of the data?
FRED (FRED, ICE, CFTC series), FMP (equities and ETFs via end-of-day), and Convex-computed composites (CVRP, CNLI, CRAI, NVI). All methodology is documented and linked from each response.
Self-host and contribute
Source is at packages/mcp-server in the Convex monorepo. Point the server at a staging or self-hosted deployment with the CONVEX_API_ORIGIN environment variable.
CONVEX_API_ORIGIN=https://staging.convextrade.com \ npx @convextrade/mcp-server