Skip to main content
MCP server CX Codex

FutureSearch MCP Server — Claude Code MCP server: forecasts, decisions

Produces probabilities, dates and numbers for questions about the future, and forecasts what each option in a decision leads to. Useful when a choice needs backing.

Install into Claude Code · Platform CLI All methods and platforms ↓
claude mcp add futuresearch-mcp -e FUTURESEARCH_API_KEY=<FUTURESEARCH_API_KEY> -- uvx futuresearch-mcp
Open on GitHub

What FutureSearch MCP Server does

The server hands your assistant forecasting tools. futuresearch_forecast covers five modes: binary probability, numeric and date percentiles, categorical, and thresholded. futuresearch_decision forecasts the same outcome under every listed alternative — what happens if a project gets one budget or another. A third tool, futuresearch_multi_agent, spins up parallel research agents across the rows of a table.

Every call is asynchronous. Starting a task returns an ID right away; you poll futuresearch_progress(task_id) for status and pull the finished output with futuresearch_results(task_id, output_path), which writes a CSV. Rows go inline as data or by reference to an uploaded artifact_id. Run the server yourself over stdio and the tools also take absolute paths to local CSV files.

Self-hosting needs a FutureSearch API key and either uv or the futuresearch-mcp package installed with pip. No key in FUTURESEARCH_API_KEY means the server will not start. The hosted endpoint at https://mcp.futuresearch.ai/mcp authenticates over OAuth and needs no key at all, which covers most people.

What it helps with

  • size up how likely an event is before committing
  • compare what one decision yields at different funding levels
  • get a date or a number as percentiles instead of a single figure
  • split a question into categories and score each one
  • run deep research down every row of a table

Example prompts for the agent

  • Forecast with futuresearch whether the team ships the release this quarter
  • Run a decision forecast on what happens at each funding level we could pick
  • Use futuresearch_multi_agent to research every row in companies.csv

How to install FutureSearch MCP Server

Platform CLI

claude mcp add futuresearch-mcp -e FUTURESEARCH_API_KEY=<FUTURESEARCH_API_KEY> -- uvx futuresearch-mcp

Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from futuresearch-mcp <command>

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about FutureSearch MCP Server

How do I install FutureSearch MCP Server in Claude Code?

Run claude mcp add futuresearch-mcp -e FUTURESEARCH_API_KEY=<FUTURESEARCH_API_KEY> -- uvx futuresearch-mcp, substituting your own key.

What does FutureSearch MCP Server do?

It gives the assistant forecasting tools: a probability, date or number for a future question, the outcomes of a decision, and parallel research across table rows.

Does the server need an API key?

For a local run, yes — the FutureSearch API key goes in FUTURESEARCH_API_KEY. The hosted server authenticates over OAuth and takes no key.

Similar MCP servers

All MCP servers for AI and ML →

From the same set2

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor