Back to directory

RS-SDK

A TypeScript SDK, agent documentation and private server emulator for building and running bots inside a RuneScape-style game, built as a testing environment for coding-agent research rather than for the live commercial game.

RS-SDK

RS-SDK is a research-oriented starter kit for building bots in a RuneScape-style game, providing a TypeScript SDK, agent-facing documentation and bindings, and its own server emulator, forked from the open-source LostCity engine and client rather than connecting to the commercial game. It is meant as a safe, bot-only setting for experimenting with goal-directed program synthesis techniques and for studying how agents collaborate and compete.

A bot connects to a provided demo server under a chosen username; high-level bot methods try to observe the actual in-game effect of an action, while lower-level methods only confirm that the client sent a request, not that the server applied it. Code can be run through an MCP execute_code tool, which receives bot and sdk objects directly, or as standalone scripts. A public leaderboard ranks bots on the demo server by account level reached per unit of playtime, and a benchmark compares different language models on the task.

It is written in TypeScript and is typically driven with a coding agent such as Claude Code.

More in AI evals

See all 269
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • JevBench

    benchmarkheaven.com

    A reproducible benchmark from Benchmark Heaven that ranks Jev-class typed decision models on intelligence, calibration, speed, and cost, with a published harness, scoring rules, and results.

  • A co-evolutionary adversarial red-teaming tool for AI agents that uses quality-diversity evolutionary search to autonomously breed edge-case scenarios, testing whether an agent with tool access can be made to hallucinate, deadlock or leak funds.