Connect your AI agent to FILTRIX

Give Claude, ChatGPT, Cursor or your own agent the FILTRIX backtest engine: historical statistics over 400K+ Trade-Ideas alerts since 2019, ticker overviews, per-ticker gap history and today's premarket gappers. Your agent calls the same API you use in the app, inside your plan and daily quota.

https://filtrix.net/mcp Remote MCP server (Streamable HTTP). Sign in with OAuth, or send an API key.

OAuth

claude.ai and Claude Desktop

Settings → Connectors → Add custom connector.
Name it FILTRIX, URL https://filtrix.net/mcp, save.
Click Connect, sign in to FILTRIX and press Allow.

Free Claude plans allow one custom connector.

OAuth

ChatGPT

Settings → Apps & Connectors → Advanced → turn on Developer mode.
Create a connector with the URL https://filtrix.net/mcp and authentication OAuth.
Sign in to FILTRIX and press Allow.

Developer mode availability depends on your ChatGPT plan and workspace settings.

OAuth or API key

Claude Code

claude mcp add --transport http filtrix https://filtrix.net/mcp

Then run /mcp in Claude Code and authenticate. With an API key instead:

claude mcp add --transport http filtrix https://filtrix.net/mcp \
  --header "X-API-Key: flt_your_key"

API key

Cursor, VS Code and other MCP clients

{
  "mcpServers": {
    "filtrix": {
      "url": "https://filtrix.net/mcp",
      "headers": { "X-API-Key": "flt_your_key" }
    }
  }
}

Any client that speaks Streamable HTTP works; Authorization: Bearer flt_your_key is accepted too.

API key

Local MCP server

Runs on your machine over stdio, same tools:

{
  "mcpServers": {
    "filtrix": {
      "command": "uvx",
      "args": ["filtrix-mcp"],
      "env": { "FILTRIX_API_KEY": "flt_your_key" }
    }
  }
}

Or from a shell: uvx filtrix-mcp.

API key

REST API and CLI

OpenAPI 3: https://filtrix.net/api/v1/openapi.json (works with GPT Actions and code generators).

curl -H "X-API-Key: flt_your_key" \
  https://filtrix.net/api/v1/meta

pipx install filtrix-cli
filtrix backtest -f "gap_gup>=30" -f "float_shares_float<10000000" --sample 5

What your agent can do

search_filters: find the right column among 250+ Trade-Ideas filters
run_backtest: what happened on the day of every matching historical event
screen_events: the matching events themselves, with any extra columns
get_ticker_overview: last close, recent moves, 52-week range, fundamentals
get_ticker_gap_stats: how one ticker's own gaps resolved
get_daily_bars: daily OHLCV
get_premarket_gappers: today's premarket movers with historical odds
get_meta / get_account: definitions, limits, your quota

Research tools work in your account, so connect with changes allowed to use them:

save_run: keep a filter set as a named run, the same runs you see in the app's Context tab
simulate_strategy: test an intraday entry and exit (the open, a color flip, a range break; the close, N minutes, a stop and target) on a run's events: win rate, profit factor, P&L per share, trades
get_trade_context: daily and minute bars of one event, to look at a trade
list_runs / get_run / share_run: your runs and their public links
get_job / cancel_job: saves and simulations run in the background
delete_run: permanent, and marked as destructive so your client asks first

From a shell the CLI does the same: filtrix runs save, filtrix simulate --wait, filtrix jobs get.

With a Pro plan, the live tools work on today's market:

scan_market: run filters over the live market right now
get_scanner_matches / save_scanner_monitor: your scanner windows and what they match
get_scanner_chart: today's minute bars and recent daily bars of one symbol
create_paper_account / save_paper_strategy / start_paper_strategy: run a strategy on paper against the live feed
get_paper_account / list_paper_records: positions, orders, trades and the journal
close_paper_run / reset_paper_account: marked as destructive so your client asks first

Paper money is simulated: nothing here places a real order. From a shell: filtrix scanner and filtrix paper.

Every number comes back with its definition and unit, so the agent does not have to guess what a statistic means.

Access and limits

Plan	Requests per day
Free	50
Pro	1,000

The day resets at midnight New York time. Definitions, limits, your account and job status checks are free calls. Saved runs and simulations also have plan limits, and at most 2 jobs run at a time. When you connect with OAuth you choose read-only or read and changes; you can disconnect an app any time in the app under API Keys, where you also create API keys.