Gamebeast Skills
Official Agent Skills and MCP server for Gamebeast, the analytics, experimentation, and LiveOps platform for games.
Connect your agent to your Gamebeast projects to query player behavior, build dashboards and cohorts, run and read out A/B experiments, tune remote configurations, and instrument your Roblox, Unity, Unreal, Godot, or web game.
Installing
Use the native plugin where supported to install both the Gamebeast skills and the Gamebeast MCP server. On first use, your client opens a browser: sign in with your Gamebeast account and choose which organizations, projects, and permissions the connection gets. Agents that only support the Agent Skills standard can install the skills separately.
Cursor
Install from the Cursor Marketplace, or add manually via Settings > Rules > Add Rule > Remote Rule (Github) with GamebeastGG/skills.
Claude Code
Install using the plugin marketplace:
/plugin marketplace add GamebeastGG/skills
/plugin install gamebeast@gamebeast
Codex
Install from the Gamebeast plugin marketplace:
codex plugin marketplace add GamebeastGG/skills
codex plugin add gamebeast@gamebeast
Start a new Codex session after installation.
VS Code / GitHub Copilot
Install directly from this repository:
- Enable
chat.plugins.enabledin VS Code settings. - Open the Command Palette and run Chat: Install Plugin From Source.
- Enter
https://github.com/GamebeastGG/skills.
For marketplace installation or troubleshooting, see VS Code's agent plugin documentation.
ChatGPT and Claude Desktop
Add Gamebeast as a custom connector using the MCP server URL https://api.gamebeast.gg/mcp. The skills are served by the MCP server itself (as skill:// resources and through the load_skill tool), so a connector-only install still gets them.
npx skills
Install using the npx skills CLI:
npx skills add https://github.com/GamebeastGG/skills
Clone / Copy
Clone this repo and copy the skill folders into the appropriate directory for your agent:
| Agent | Skill Directory | Docs |
|---|---|---|
| Claude Code | ~/.claude/skills/ | docs |
| Cursor | ~/.cursor/skills/ | docs |
| OpenCode | ~/.config/opencode/skills/ | docs |
| OpenAI Codex | ~/.codex/skills/ | docs |
Then add the MCP server yourself (see MCP Server).
Skills
Skills are contextual and auto-loaded based on your conversation. When a request matches a skill's triggers, the agent loads and applies it.
Playbooks
Multi-step jobs. Each is an approach the agent adapts to what you asked, not a fixed script.
| Skill | Useful for |
|---|---|
| instrument-game | Designing a tracking plan and remote configuration, wiring the SDK calls, and verifying each event against live data |
| build-team-dashboards | Building a project's first engagement, retention, economy, and monetization dashboards from the events it actually sends |
| weekly-review | A two-minute weekly summary: what moved and why, running experiments, player sentiment, and what to do next |
| diagnose-change | Investigating why a metric moved or an event stopped, down to the segment, config edit, or experiment behind it |
| fix-instrumentation | Auditing event data for stopped events, drifting property types, high cardinality, and tracking-plan mismatches |
| ship-or-kill-experiments | Deciding running experiments (ship, end, extend, or ramp) with sample-ratio, power, and duration checks |
| launch-liveops-event | Running a measurable LiveOps event as a holdout experiment with a staged rollout, KPI dashboard, and readout |
| tune-economy | Finding and fixing in-game economy imbalances and shipping the fix as a reversible experiment |
| target-segment | Finding the players a change should reach, as a cohort, and targeting an experiment at it |
| player-voice-to-action | Turning community feedback into a testable change backed by behavioral data |
References
| Skill | Useful for |
|---|---|
| insight-queries | Insight specs: metrics, aggregations, formulas, filters, breakdowns, and value formatting |
| funnel-queries | Funnel specs: steps, conversion windows, ordering, and breakdowns |
| retention-queries | Retention specs: start and return events, day-N / week-N windows, and displays |
| heatmap-queries | Heatmaps and heatmap queries: counting an event's positions on a map and reading hotspots |
| run-analytics-query | Running saved queries with time or breakdown overrides and interpreting the results |
| cohorts | Reading and building cohort (player segment) definitions |
| manage-cohorts | Creating, editing, and deleting saved cohorts |
| manage-dashboards | Creating, laying out, organizing, and deleting dashboards and saved queries |
| manage-events | Describing analytics events and their arguments, and hiding or showing them |
| understand-experiments | The experiment model and how to measure and read results |
| manage-experiments | Designing, previewing, creating, ending, and assigning A/B experiments |
| experiment-rollouts | Staged rollouts: ramp curves, rollout steps, and operating a ramp |
| understand-configurations | Navigating, reading, and searching remote configurations |
| understand-deeplinks | Deeplink definitions and their interaction, conversion, and LTV analytics |
| understand-community-pulse | Community Pulse: sentiment sources, topic trends, complaints, praise, and quotes |
MCP Server
This plugin includes the Gamebeast remote MCP server:
| Server | URL | Purpose |
|---|---|---|
| gamebeast | https://api.gamebeast.gg/mcp | Query analytics, manage dashboards, cohorts, configurations, and experiments, and read community sentiment across your Gamebeast projects |
The server signs you in with OAuth; there's no token to copy. The connection is read-only apart from configurations, experiments, and analytics assets (dashboards, saved queries, cohorts, event descriptions), and only if you grant write permissions when connecting. Your own role is re-checked on every request, so the agent can never do more than you can. MCP calls don't use your Gamebeast AI credits.
For personal access tokens, per-project URLs, and the full permission model, see the MCP server docs.