Model Context Protocol
Connect your AI assistant to Feedback in Games. Explore what players are saying, draft better surveys, and bring your content to more languages.
Your account. Your games. Permissions you approve.
A focused analysis workflow
Find the right survey
Feedback in Gamessearch_surveysCompare response cohorts
get_survey_analyticsInspect the supporting answers
get_survey_responsesUse natural language. Keep the results grounded in your data.
01 / Connect
Use your existing Feedback in Games account. Your assistant connects through OAuth, so you can choose its permissions and revoke access later.
Start with analysis, or select the work you want to do.
Toolsets filter the catalog. You choose permissions separately when you sign in. Keep the query parameter when adding a focused connection.
Add the server from your terminal, then open Claude Code.
claude mcp add --transport http feedback-in-games 'https://app.feedbackingames.com/api/mcp?toolsets=analysis'Run /mcp, select the server and authenticate. Sign in to Feedback in Games and approve the permissions you need.
“Show my accessible games and the surveys in my game.”
Setup guidance checked against official client documentation on October 11, 2026. Available setup options depend on your client and workspace. See compatibility and verification.
02 / Put it to work
These prompts are starting points. Your assistant can look up IDs, inspect the current state and use the right tools for each step.
Compare response cohorts, inspect distributions and follow up on individual answers.
Find my onboarding survey. Summarize its responses and compare completion and question results between two app versions. Explain the denominators and any limitations.
search_gamessearch_surveysget_survey_analyticsget_survey_responsesCreate a draft with questions and rewards, then review readiness before an explicit activation.
Draft a three-question survey about the first level of my game. Keep it as a draft, validate it, and show me any issues before we decide to activate it.
create_survey_from_specvalidate_surveyupdate_surveyInspect source text and existing translations, then apply a batch of reviewed changes.
Review my survey's Spanish translations. Show the missing text and propose translations. Apply them only after I review the changes.
list_survey_translationsupsert_question_translationvalidate_surveyInspect targeting, cooldowns and availability without serving a survey or charging credits.
Help me understand why a player is not eligible for my survey. Ask me for the player ID and request context, then explain every blocking condition without changing anything.
get_surveyexplain_survey_eligibility03 / Access you control
mcp:readInspect games, surveys, responses, KPIs and translations. Start here for analysis.
mcp:writeDraft and update game content, survey settings and translations. Add read access when the workflow also inspects data.
mcp:adminUse profile, team and billing operations. Administration requires an all-games connection; team roles still apply.
Scopes are independent. A grant never overrides ownership, selected-game access, team roles or plan requirements. Toolsets organize tools; they are not security boundaries.
Choose selected games during consent, or allow all accessible games. Reconnect to change your grant. Manage MCP connections to revoke access; new requests and refreshes are blocked immediately. Work already in progress may finish.
04 / Tool reference
Explore the complete catalog, generated from the server's tool definitions. Your connected client sees the tools allowed by its toolsets, scopes and game policy. Open a tool for its behavior and input schema.
97 of 97 tools · showing 12
Full JSON reference ↗Successful calls return structuredContent.data; failures return structuredContent.error with a code, message and recovery guidance. Compatible text content is also returned for older clients. SDK input and transport errors use the protocol's standard format.
The JSON reference includes the shared output schema. Individual result payloads are not exhaustively specified by that envelope. Server-side refinements and business checks still apply to the documented inputs.
05 / Good to know
Complete the client's OAuth sign-in flow. Tokens expire and revoked connections cannot be reused. Reconnect if the grant was removed; an old game SDK key will not authenticate an MCP connection.
Check the connection's toolset, approved scopes and selected games. The reference shows the full catalog, while your client sees only permitted tools. Account-level operations require all-games access. Reconnect to change permissions; business ownership and team roles still apply.
Read permission does not include write or administration. Approve the required scope during reconnection. Team and billing actions may also require a team-admin role, and some capabilities require a plan entitlement.
Use a smaller page or narrower filters and follow next_page or next_cursor. Some tools have independent pages for each collection. Prefer get_survey_analytics for aggregates. Oversized results fail explicitly; never treat one page as a complete editable definition.
The server accepts configured browser origins. Ask the server operator to allow your client's exact origin using MCP_ALLOWED_ORIGINS. CLI clients may omit Origin. Do not disable origin validation.
The /api/mcp URL is an authenticated protocol endpoint, not a web page. Add it to your MCP client's remote-server settings. Read these instructions at /mcp. Use Streamable HTTP; legacy HTTP+SSE endpoints are not provided.
Streamable HTTP supports MCP 2026-07-28, with compatibility for 2025-11-25, 2025-06-18, 2025-03-26. The server is stateless and does not provide legacy HTTP+SSE endpoints.
Protocol and authorization behavior is covered by automated server and official TypeScript SDK tests. Native client sign-in and tool-loading behavior have not been verified. Client-side discovery and context savings depend on the host.