Choose theme
AI Resources
Horizon
Horizon is an unofficial Rust reimplementation of Minecraft: Java Edition with a control interface for AI agents and scripts. A connected agent can read the game state and take player actions through a command-line tool or MCP server.
The project includes a playable client, launcher and server. Its AI relevance is the ability to interact with that game through structured tools: observe the world, move, mine, craft and examine what happened. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
A playable world with agent controls
The horizon-ctl command-line tool exposes player state, inventory, blocks, entities and menus as JSON. It can also move the player, use items and interact with the world. An agent supplies the decisions; Horizon supplies the game and the controls.
Why it stands out
Game actions return their outcomes
The documented commands wait for an action to finish or time out. A movement command reports whether the player arrived; mining reports drops and pickups. This gives a script an action result to work with before it chooses the next step.
Availability
Source, launcher and platform packages
The repository publishes source and release packages, with a launcher for Linux x86-64, Apple-silicon macOS and Windows x64. It declares an MIT licence; see the main repository for current terms. Minecraft installation files are a separate requirement.
Why it matters
What makes it useful
A builder can use the same control interface for a shell script or an MCP-connected agent. The documented example starts a world, finds a tree, walks to it, mines a log, crafts planks and places a block. State queries include health, hunger, position and the targeted block, while inventory queries expose the materials on hand. These are concrete inputs and actions for experimenting with an agent that plays a game.
What to know
Where it fits
Install and prepare the game through the Horizon launcher, then use horizon-ctl to launch a world with control enabled. For MCP, the native-client guide documents registering horizon-ctl with the argument mcp as a stdio server. It exposes one tool per command, and screenshot results are images. The client or agent that chooses those tools remains a separate part of the setup.
Notable points
What stands out
The interface includes events and wait-for commands as well as immediate observations. A script can wait for a screen, widget, chat message or event instead of repeatedly querying the game. Events carry sequence numbers and game ticks. The guide shows taking the latest sequence before an action, then waiting for a later event so an event that happens quickly is not missed.
Before using
What to review
Horizon is not an official Minecraft product and is not associated with Mojang or Microsoft. Its documented preparation uses your own Minecraft: Java Edition 26.3 installation and Java 25. The launcher and installation guide describe the platform and graphics requirements.
The project lists Microsoft-account sign-in as missing, which prevents joining servers that require account authentication. Java mod jars do not run; Horizon has its own mod system. These affect whether it fits an existing server or mod setup.
In agent-control mode, losing window focus does not pause the game or release held keys. The control guide documents stop, pause and quit commands alongside movement and interaction tools; this matters when running a script while using another window.
Reader fit
Who may find it relevant
Agent builders, automation developers and technically comfortable Minecraft players are the clearest audience. The command reference and source offer a way to explore a real game interface rather than only read about agent workflows. Someone mainly looking to play on an existing authenticated server or use a collection of Java mods should review those compatibility gaps first.
Editorial note
Why LifeHubber lists it
LifeHubber lists Horizon because its structured controls put agent decisions inside a world with connected tasks. Crafting depends on having materials, and gathering those materials involves movement and interaction. For a builder exploring multi-step agents, the game makes those dependencies tangible: inventory and action results show how one step changes what can happen next.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
More in AI Agents
Keep browsing this category
Explore more AI agent projects.
Paperclip
paperclipai/paperclip
A self-hosted server and dashboard for coordinating agent teams through companies, goals, roles, issues, heartbeats, budgets, approvals, and persistent activity records.
Tracely
jwuthri-cloud/Tracely-ai
A self-hostable agent tracing and regression-testing toolkit that connects failed runs, recorded test cases, replay results and CI checks.
Headlong
laude-institute/headlong
A Laude Institute agent microharness whose Bash-based runtime keeps one agent thinking between messages, records its trajectory in inspectable files, and connects the same thought stream to terminal, web, Slack, or Telegram.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
