Skip to content
Game Market Copilot Docs
Esc
↑↓navigate↵open⌘Jpreview
On this page

MCP servers for GMC

How to connect an agent to the GMC market-data MCP server and the read-only Docs MCP, including tool list, cost classes, and setup commands.

Game Market Copilot exposes two separate MCP servers. They serve different purposes and should not be confused.

Which server to use

Server Endpoint Auth Serves Billing
gmc-docs (this site) https://docs.gamemarketcopilot.com/mcp none Read-only product documentation: setup guides, API reference pages, this methodology page none, no credits
gmc (market data) https://gamemarketcopilot.com/api/mcp/mcp OAuth (auto-discovered) or bearer API key Steam market analytics: search, aggregates, cohort evidence, showcase matching, game lists Credit-charged; requires a paid plan

Use gmc-docs for product, setup, and how-to questions. Use the gmc MCP for market-data operations. If an agent needs both, add both servers.

Every page on this site is also available as raw Markdown by appending .md to its URL, and as consolidated llms.txt / llms-full.txt files, independent of the MCP connection.

Setting up the gmc-docs MCP

No account or credentials are required.

Claude Code

claude mcp add --scope user --transport http gmc-docs https://docs.gamemarketcopilot.com/mcp

Codex

codex mcp add gmc-docs --url https://docs.gamemarketcopilot.com/mcp

Available tools: search_docs, get_page, list_pages, get_navigation.

Setting up the gmc MCP (market data)

Requires a paid GMC plan (Starter or above). Free workspaces cannot mint MCP credentials.

Claude Code

claude mcp add --scope user --transport http gmc https://gamemarketcopilot.com/api/mcp/mcp

Or install through the plugin marketplace:

/plugin marketplace add witchpot-studio/gmc-agent-skills
/plugin install gmc@gmc

Codex

codex mcp add gmc --url https://gamemarketcopilot.com/api/mcp/mcp
codex mcp login gmc

The server advertises OAuth discovery metadata, so compliant clients can authenticate interactively without extra configuration. For headless setups, a bearer API key works in place of OAuth.

Transport is Streamable HTTP.

Tools

The gmc MCP publishes 16 tools. Costs are described in classes, not exact figures. The precise charge for any call appears in that call’s response metadata (meta.credits_charged, meta.quota) and on the workspace usage screen.

Tool Purpose Cost class
coverage_check Check how much of a cohort has been researched before fanning out per-title calls baseline
resolve Resolve a free-text name to a canonical appid baseline
list_games Search or list titles by filter baseline
market_aggregate Population-level aggregates (counts, distributions, group-bys) high; count-only queries are cheaper than distribution queries
game_profile Per-title profile (detail, reviews, marketing, creators, success report) scales with the number of titles and the depth of sections requested
cohort_evidence Sampled review evidence across a cohort of titles scales with the number of titles sampled
cohort_review_categories Cohort-level review theme synthesis (population-true rollup) high
search_review_claims Per-title search over named review claims mid
showcase_fit Match a title against festivals and showcases low-mid
Game list tools (list_game_lists, get_game_list, create_game_list, update_game_list, add_game_list_items, remove_game_list_item, delete_game_list) Manage saved, reusable cohort definitions free (no credits), but still requires a paid plan

General rules that apply across tools:

  • Prefer market_aggregate over paginating list_games for sizing and distribution questions. For population-wide questions it usually costs less in total than paging through results, and it returns population-true numbers instead of a page-bounded sample.
  • Call coverage_check before fanning out game_profile across more than a few titles.
  • not_collected means the data was not collected. It is never equivalent to zero events or absent sentiment.
  • Figures not derived from GMC data (unit multipliers, wishlist folklore, revenue guesses) must be labeled as external estimates, never blended into GMC-derived numbers.

See /methodology for how coverage, sampling, and tag-counting bases affect what a response means.

Guided prompts

The gmc MCP also serves four guided prompts that embed the analysis guardrails and an output contract, so a client does not need the gmc-analysis Skill installed to get a validated workflow:

  • cohort_complaints: what players dislike or love about a cohort of titles
  • market_sizing: how big a market segment is and how it is distributed
  • title_deep_dive: why a specific title performed the way it did
  • design_pillars: design-pillar candidates derived from comparable-title evidence

Prompts are static text only; they do not call the server or charge credits themselves. They tell the client which tools to call and in what order. For agents that can run shell commands or read a repository, the gmc-analysis Skill covers the same workflows with more depth, including chart-card guidance.

Was this page helpful?