A finance MCP server — market and macro-economic data for AI agents.
MacroView is a finance MCP server: 51 tools that plug live market data and macro-economic insights straight into any AI agent over the Model Context Protocol, all from the same feed the MacroView dashboard runs on.
Market and theme heatmaps, a volatility regime read, carry-unwind and gold-driver gauges. A calendar of IPOs, earnings, FOMC/CPI/jobs releases, options expiry, index rebalances and dividends, with post-earnings reaction history behind it and an earnings wire of SEC 8-Ks caught minutes after filing. Six screeners over the US tape: 52-week lows, RSI tails, the names that just crossed a line, weekly movers, an attention rank, and a rotation board of where price leadership went by ETF and the names carrying it. Per-ticker support & resistance levels with the touch history behind every line, and the 50/100/200-day moving-average configuration of the major US and European indexes read against their region’s corporate-profit trend.
On the filings side: 13F holdings diffs, a disclosure wire of 13D/13G activist stakes and Form 4 insider buying, the UK short register and futures positioning. Macro adds a sovereign-debt network, an AI build-out financing map reading the hyperscalers’ funding mix straight out of SEC filings, a cash-and-gold capital-parking picker and a daily news digest. Plus a 100k paper-trading account your agent can actually place simulated orders in.
Every tool advertises a one-or-two-sentence summary saying which question it answers and which neighbouring tool to prefer instead, plus its full input schema — so the catalog costs your agent a quarter of the context it used to. One describe_dataset call returns the complete documentation — caveats, units, publication lags — for whichever tool it picks.
One endpoint. Every client.
Add MacroView as a remote MCP server. Claude Code, Codex, Cursor, VS Code and Gemini CLI sign in through the browser and refresh their own token, so those snippets carry no key at all — only opencode, pi and the OpenAI Responses API need the placeholder replaced with your bearer token.
"tools": [{ "type": "mcp", "server_label": "macroview", "server_url": "https://macro-view.com/mcp", "authorization": "etfffs_sk_YOUR_KEY", "require_approval": "never" }]
What is a finance MCP server?
The Model Context Protocol is an open standard for letting a large language model (LLM) discover and call tools it was never specifically programmed for. A finance MCP server is one that exposes market and macro-economic datasets as those tools — so an agent asks for the volatility regime or the earnings calendar the same way it would call any other capability, without anyone writing an integration for it first.
That is the difference from handing a model a REST API key. An API is a schema someone has to teach it; a catalogue is something it can enumerate. Each of MacroView’s 51 tools advertises what its dataset is and when to reach for it, so the model plans with the data instead of guessing at endpoints — which is also what makes the same server usable for quantitative research, a trading-desk chat assistant or a scheduled research routine without changing anything on this side.
One honest caveat, because it decides whether this is the right tool for your job: this is not a tick feed. Everything serves from a 30 min – 12 h server cache, and every payload carries the timestamp it was cached at plus its remaining TTL, so an agent can always tell fresh from stale rather than presenting an hour-old quote as live. For macro regimes, calendars, screeners and filings that is the natural cadence of the data. For execution or intraday signals, it is not, and you want a market-data feed instead.
All 51 tools
A complete mirror of the live datasets behind the dashboard, plus one tool that documents the rest and one that reports a gap back to us. Every data tool returns structured JSON with an updatedAt and ttl so an agent knows how fresh the data is. The data tools only read; of the 7 marked writes, four place simulated orders in your own paper-trading account, two let a Quant or Teams admin run their tenant, and one files a gap report when the data can’t answer the question.
Example prompts
FAQ
What is the MacroView MCP server?
MacroView is a finance MCP server with 51 tools, spoken over the Model Context Protocol at macro-view.com/mcp. The read-only half covers market and sector heatmaps, curated theme baskets running from the semiconductor complex to energy and power, AI infrastructure, biotech, optics, cybersecurity, quantum, space, drones and robotics, volatility and macro gauges, and a market-event calendar of IPOs, earnings, FOMC/CPI/jobs releases, options expiry, index rebalances and dividends with post-earnings reaction history behind it. An earnings wire picks up SEC 8-Ks within minutes of filing. Six screeners run across the US tape: 52-week lows, an oversold/overbought RSI screen over every $2B+ listing, a breakout screen of the names that just crossed an RSI line, a moving average or an EMA pair with candle and volume confirmation by sector, weekly movers above 6%, an attention rank, and a rotation board of 60 ETFs wide to narrow with the $10B+ names carrying each move. Per-ticker support and resistance levels come with the touch history behind every line, and the 50/100/200-day moving-average configuration of the major US and European indexes is read against each region's corporate-profit trend. On the disclosure side: quarterly 13F holdings diffs, 13D/13G activist stakes, Form 4 insider buying market-wide, a per-ticker insider ledger that keeps open-market buying apart from pre-arranged plan selling, the FCA's UK short register and CFTC futures positioning. Macro adds a sovereign-debt network, an AI build-out financing map that reads the hyperscalers' capex funding mix out of SEC filings and sets it against credit spreads, a cash-and-gold capital-parking picker, ranked gold, silver and copper miner ETF shortlists for EU and US buyers, and a daily news digest. Five further tools let an agent run your paper-trading account and three let a Quant or Teams admin run their tenant, covering seats, per-member request budgets and service keys. One tool returns the full documentation for any other, so an agent's tool list stays small and every caveat is one call away, and one lets an agent tell us what it could not answer.
Can an AI agent place trades?
Only simulated ones. On the Analyst plan and up, an agent can read your investment simulator, open one, place buy and sell orders (buys can carry a stop loss and take profit), set exit levels on a position, and reset the account — a fictional 100,000 EUR or USD account, filled at MacroView's cached quotes. No order ever reaches a broker and no real money is involved. It is the same account the dashboard shows, so an agent's trades appear there too, and the tools are marked as writes so your client can ask you before running them.
How do stop-loss and take-profit work in the simulator?
You attach them to a position, from the dashboard or through an agent, as prices in the instrument's own currency. They are settled when the account is read rather than watched tick by tick: each read scans the daily closes since you set the levels, and the first close through a level closes the whole position at that close's price and date. They are also the only way to sell at a price — an order itself always fills at the market, right away. Intraday spikes, and a level crossed and recovered between two closes, are missed by design — the whole simulator runs on cached prices, so a bracket is a realistic stop, not a resting order on an exchange.
Does the simulator model leverage on FX?
Yes. FX pairs trade on 1:30 margin, the retail default, so 3.33% of the position is posted as cash and the profit and loss runs on the whole notional — which is what makes a 0.5% currency move visible in a 100,000 account. Stocks, ETFs and indices are unlevered. A levered position that loses more than the margin behind it is closed automatically, and every position reports its margin, notional and return on the cash committed.
Which AI agents and clients can connect?
Any MCP-capable client: Claude Code, Claude Desktop and claude.ai, OpenAI Codex, Cursor, VS Code and GitHub Copilot, Gemini CLI, opencode and pi, plus the OpenAI Responses API. Copy-paste configuration for each is on this page.
Do I need an API key?
Not for most clients. Claude Code, Claude Desktop, claude.ai, OpenAI Codex, Cursor, VS Code and Gemini CLI all speak the MCP OAuth 2.1 flow: point them at the endpoint, approve the sign-in in your browser once, and they refresh their own token from then on — nothing to paste, nothing to rotate. A per-user API key (etfffs_sk_…) is still available for callers that have no browser, such as the OpenAI Responses API, the Anthropic MCP connector and CI, and for direct REST calls. Either way the request needs an active plan. The REST API behind the dashboard is session-gated for signed-in browsing; only the sectors heatmap, Mag 7 and earnings-bubble teasers are keyless.
Is the data real-time? How fresh is it?
Not real-time in the tick-feed sense, and the distinction decides whether this is the right tool for your job: it suits macro regimes, calendars, screeners, filings and positioning, not execution or intraday signals. Tools serve from a 30 min – 12 h server cache — 30 minutes on the stock cards, the sectors heatmap, the 52-week-low and RSI screens and price alerts, 60 on the macro sets (3 hours on weekends; the hot-stocks scan and gold drivers refresh every 6 hours and the thematic and regional heatmaps every 12, because a week of traffic said nobody was reading them often enough to warm harder; the sovereign-debt network, the moving-average trend read and futures positioning refresh every 12 hours, the 13F holdings diffs every 6; the 13D/13G and Form 4 disclosure feeds every 30 minutes, the UK short register once each working day; the earnings wire is fed within minutes of an earnings 8-K landing on SEC EDGAR). Every payload carries updatedAt and ttl, so an agent knows exactly how stale the data is.
Can I use it for quantitative research?
Within the cadence above, yes — that is much of what it is for. Price history comes back as raw [timestamp, value] pairs on request, screeners run across every $2B+ US listing, and the filings wires (13D/13G, Form 4, the UK short register, 8-K earnings) are the raw disclosure events rather than a summary of them. It is a research and monitoring surface, not a backtesting warehouse: history depth varies by dataset, and there is no point-in-time reconstruction of what a figure looked like before it was revised.
What markets does it cover?
Equities and ETFs, sectors, forex, energy, bonds and sovereign debt, and gold, the last including money-market and short-bill funds and physical gold ETCs with their fees, venues and availability. The AI-infrastructure complex is covered on three sides: the semiconductor chain; the cloud hyperscalers, neocloud GPU providers and data-center operators that host the compute; and how the build-out is financed, through the vendor-financing circle, capex versus operating cash flow straight from SEC filings, and the credit spreads behind it. Beside it sit the listed cybersecurity complex, developer tools and infrastructure, and the quantum-computing field down to a $2B market cap. Volatility comes through VIX, MOVE and SKEW. The market-event calendar carries IPOs, large-cap earnings, FOMC/CPI/jobs release dates, options expiry and quad witching, index rebalance schedules, ex-dividends and IPO lockup expiries, with post-earnings reaction history and an earnings wire of just-filed 8-Ks. Screeners run across the US tape: 52-week lows among $20B+ large caps, an RSI(14) oversold/overbought screen over every $2B+ listing, and a breakout screen of the names whose settled close just crossed the RSI 30/70 lines, the 9/20 EMAs or the 50/150-day averages with candle and volume confirmation. Per-ticker support and resistance levels span windows from six months to five years, and the 50/100/200-day configuration of the S&P 500, Nasdaq 100, STOXX Europe 600 and EURO STOXX 50 is read against US after-tax corporate profits and euro-area gross operating surplus. Filings and positioning close it out: quarterly 13F holdings diffs, SEC Schedule 13D/13G activist stakes and Form 4 insider purchases as they are filed, a per-ticker insider ledger covering a year of one issuer's buying and selling with 10b5-1 plan sales split out, the FCA's aggregated UK short register, CFTC futures positioning, and a clustered daily news digest.
Is MacroView free?
The MacroView dashboard is free to explore. Agent access over MCP needs a paid plan, whether the client signs in through the browser or sends an API key — pick a plan in the app and the connection works immediately.