Skip to main content
Glama
steamforecast-mcp

steamforecast-mcp

by GC108

steamforecast-mcp

CI PyPI License: MIT

Model Context Protocol server for SteamForecast, the Steam launch revenue forecaster. Exposes calibrated revenue cones (P10–P90; ~82% nominal coverage, 81–86% realized per wishlist tier on 6,422 held-out launches) to Claude, ChatGPT, and any MCP-aware AI agent as tool calls.

What it does

Five tools, all backed by the public steamforecast.app API:

Tool

What it does

get_forecast(appid)

Calibrated P10/P50/P90 revenue cone for a Steam game by appid

get_comps(appid, k)

Top-K nearest-neighbor comparable games (cosine sim over BGE embeddings)

boxleiter_estimate(review_count, price_cents)

Pure-compute Boxleiter rule-of-thumb sanity check

get_calibration_summary()

Latest published live coverage table (per-stratum)

get_methodology()

Pulls llms.txt — high-quality URL inventory for ingestion

get_forecast and get_comps make HTTPS calls to steamforecast.app. The other three are pure compute / static reference, so they work offline once the package is installed.

Related MCP server: OpenDota MCP Server

Install

pip install steamforecast-mcp

Configure your MCP client

Claude Desktop / Claude Code

Add to your MCP config (typically ~/.claude.json or ~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "steamforecast": {
      "command": "steamforecast-mcp"
    }
  }
}

Or via the Claude Code CLI:

claude mcp add steamforecast -- steamforecast-mcp

Other MCP clients (Cursor, Cline, etc.)

Use the standard stdio MCP config; the executable is steamforecast-mcp and takes no arguments.

Quick usage

Once configured, ask your AI agent things like:

"Pull a calibrated revenue forecast for Hades on Steam (appid 1145360) and compare it to the Boxleiter rule of thumb. Are they consistent?"

The agent will call get_forecast(1145360), then call boxleiter_estimate(review_count, price_cents) with values from the forecast result, then surface the divergence to you.

"What's the live calibration coverage on the strategy_sim stratum?"

The agent calls get_calibration_summary() and reads the per_stratum table.

Why a separate server when the website exists?

Because LLMs and AI agents shouldn't have to scrape HTML to use a calibrated forecast. The MCP surface is structured (typed JSON), versioned, and rate-limit-aware, which is the right contract for tool-using models.

It also lets you build automations without manually copying numbers from the website into spreadsheets — e.g., a nightly Claude Code routine that pulls a forecast for every appid in a publisher's portfolio and writes a report.

Configuration

Env var

Purpose

Default

STEAMFORECAST_BASE_URL

Override the API base URL (useful for local dev / staging)

https://steamforecast.app

Development

git clone https://github.com/GC108/steamforecast-mcp
cd steamforecast-mcp
pip install -e ".[dev]"
pytest
ruff check .

License

MIT — see LICENSE.

Available Tools

5 tools
boxleiter_estimateA

Apply the Boxleiter rule-of-thumb (review_count × multiplier × price).

A heuristic sanity check, NOT a calibrated forecast. Per the formula's own author (Mike Boxleiter, 2023 retrospective), ~24% of games are off by more than 30% from a single-multiplier estimate. Useful to compare against get_forecast() — large divergence between the heuristic and the calibrated cone signals an interesting outlier worth investigating.

Args: review_count: Total Steam reviews on the game's page. price_cents: List price in cents (e.g. 2499 for $24.99).

Returns: Dict with low (×30) / median (×50) / high (×63) revenue brackets in cents + dollars + a calibration warning.

Inputs: review_count (required), price_cents (required). Returns: object

get_calibration_summaryA

Return the latest published live calibration coverage summary.

Numbers are from the Q2 2026 quarterly report. Live-refreshed table is at https://steamforecast.app/methodology — fetch get_methodology() for the canonical current values.

Returns: Dict with aggregate coverage, per-stratum coverage table, sample sizes, and link to the live page + quarterly report.

Inputs: no named parameters. Returns: object

get_compsA

Fetch top-K nearest-neighbor comparable Steam games for an appid.

Comps are surfaced via pgvector cosine-similarity over a 1024-dim BGE embedding of game metadata (genres, tags, language, platform support, multiplayer features). Useful for sanity-checking a forecast: if the nearest comps cluster in a tight revenue band, the cone is likely well-anchored; if they're dispersed, the cone correctly widens.

Args: appid: Steam app ID to find comps for. k: Number of comps to return (1-20, default 5).

Returns: Dict with appid + list of comps, each including release year, price, follower count, week-1 + lifetime revenue, cosine similarity.

Inputs: appid (required), k. Returns: object

get_forecastA

Fetch a calibrated P10–P90 revenue cone for a Steam game by appid.

Uses the same v1.1 model that powers the public steamforecast.app site. Returns a JSON object with cone bounds in cents + dollars, the model version, the genre cluster used for stratified calibration, and links back to the methodology page + latest calibration report.

Args: appid: Steam app ID (e.g. 1145360 for Hades). wishlist: Optional override for catalog wishlist count (what-if mode). followers: Optional override for catalog SteamCommunity follower count.

Returns: Dict with appid, name, genres, p10/p50/p90 revenue, methodology URL.

Raises: httpx.HTTPStatusError: 404 if appid not in v1.1 catalog (~49K apps); 503 if forecast model is briefly unloaded during a deploy.

Inputs: appid (required), wishlist, followers. Returns: object

get_methodologyA

Return the AI-crawler-friendly methodology summary (llms.txt).

Pulls the canonical content discovery file from steamforecast.app/llms.txt, which lists high-quality URLs (methodology, guides, reports, tools) for AI agents to ingest. Useful when a model wants the full sitemap of authoritative content rather than a single forecast.

Returns: Plaintext content of /llms.txt (markdown-formatted per llmstxt.org).

Inputs: no named parameters. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.6/5.0

    Scored across 5 tools

    Disambiguation5/5

    Each tool has a distinct purpose: heuristic estimation, calibrated forecast, calibration summary, comparable games, and methodology. No overlap or ambiguity.

    Naming Consistency4/5

    Four of five tools follow the 'get_<noun>' pattern. The outlier 'boxleiter_estimate' uses a different convention, but it is still descriptive and not confusing.

    Tool Count5/5

    Five tools is well-scoped for a forecasting server, covering the main forecast, heuristic, comps, calibration, and methodology without excess or deficiency.

    Completeness5/5

    The tool surface covers all key operations for the domain: generating forecasts, sanity checks, comparison, calibration context, and documentation. No obvious gaps.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers