Plan an FTC autonomous on the real field, see whether it will work, and run it on the robot without editing Java.

Zenith is an autonomous path planner for FTC. You draw a routine on the field as paths with headings, attach robot commands to points on those paths, and check the plan against the field, period and robot limits in your files. AI agents edit the same files through a CLI and an MCP server and get the same checks. Each routine is one readable JSON file in your robot repository, and a small runtime on the robot plays it through Pedro Pathing v3 and Ivy or SolversLib. Once that runtime and your named commands are set up, a new auto needs no Java edit.

Zenith also plans with your alliance partner. Load a partner's file next to your auto to see both robots on one field and one clock, with predicted collisions and one-click fixes; declare the area a step that paths on the fly from vision stays in; and suggest changes to a partner's auto as a file they accept or reject change by change. Paths come in from Pedro Pathing Visualizer projects, RoadRunner through MeepMeep, and two documented interchange files. A web app on the editor's allowed list, such as the DSIM simulator, can open Zenith in a popup with its own robot project and take the saved auto back.

Zenith is in beta. It works end to end, but file formats, command names and the editor layout may still change before 1.0. Pin the version you use and read the changelog before you upgrade.

Start here

	What it is	Where
Desktop	The Windows app. Opens a robot project folder in place, and adds the high-fidelity sim, deploy, commit and pull request buttons.	GitHub Releases
Web	The same editor in a browser. Nothing to install; the example loads with one click.	zenith.horizon36596.org
Agents	The zenith CLI and the zenith-mcp server, for Claude Code or any MCP client.	Agents and MCP
Robot runtime	The Java library that plays an auto file on the robot.	Robot runtime

The documentation is at https://libraries.horizon36596.org/zenith/: getting started, the editor guide, a newcomer's guide to autonomous paths, the file-format reference and the JSON Schemas, the checks, the sims, alliance partners, imports from other planners, host mode, the CLI, agents and MCP, seasons and the FAQ.

Accounts, teams and sharing

Zenith works fully with no account. Optional accounts add a team library in the cloud and sharing with other teams, encrypted end to end on your devices. They are in the published web app and the desktop app: anyone can sign in with GitHub, Microsoft or Google and make or join a team. Teams and sharing describes them, and the privacy notice says what the cloud holds. A build from source has the cloud off unless it is given a Firebase config (cloud/README.md).

Works with SimLoop

SimLoop is Horizon's library for running FTC robot code on a laptop: your subsystems and OpMode run against fake hardware in a JVM unit test. Zenith works with it. Zenith's high-fidelity sim runs your robot repository's own headless sim, and a SimLoop setup can be that sim. Zenith does not depend on SimLoop, and nothing in Zenith needs it to plan, check or preview an auto. SimLoop is an alpha, v0.2.0-alpha.1, under the BSD-3-Clause licence.

Full sim with SimLoop sets the two up together, and the SimLoop docs cover SimLoop itself.

Install

The CLI and the MCP server, from npm:

npm install --save-dev --save-exact @horizon36596/zenith-cli @horizon36596/zenith-mcp

Or point your MCP client at the server with no install, in .mcp.json at your robot repository's root:

{
  "mcpServers": {
    "zenith": { "command": "npx", "args": ["-y", "@horizon36596/zenith-mcp"] }
  }
}

The robot runtime, from JitPack, in build.dependencies.gradle at the root of your FTC project. It runs on Ivy, Pedro Pathing's command library, or on SolversLib; both are first-class, and you declare the one your robot code uses.

For Ivy:

repositories {
    mavenCentral()
    maven { url = 'https://jitpack.io' }                                       // Zenith
    maven { url = 'https://central.sonatype.com/repository/maven-snapshots' }  // Pedro Pathing v3
}

dependencies {
    implementation 'com.pedropathing.ivy:pedro:1.1.1'
    implementation 'com.pedropathing:core:3.0.0-20260828.185437-17'
    implementation 'com.github.Horizon-36596:zenith:v0.1.2'
}

For SolversLib:

repositories {
    maven { url = 'https://jitpack.io' }                                       // Zenith
    maven { url = 'https://repo.dairy.foundation/releases' }                   // SolversLib
    maven { url = 'https://central.sonatype.com/repository/maven-snapshots' }  // Pedro Pathing v3
}

dependencies {
    implementation 'org.solverslib:core:0.3.6'
    implementation 'com.pedropathing:core:3.0.0-20260828.185437-17'
    implementation 'com.github.Horizon-36596:zenith:v0.1.2'
}

Then name the same library in the deploy block of zenith.json, "commandLibrary": "ivy" or "commandLibrary": "solverslib". Zenith does not pick one, and writes no Java until it is set. The robot runtime page covers the Pedro and command library setup and deploying autos, and Choosing a command library says how to choose and how to switch. The SolversLib runtime is the one tested most on a real robot.

Build it yourself

You need Node 22 and pnpm 10.

pnpm install; pnpm build; pnpm test

pnpm --filter ./apps/web dev

CONTRIBUTING.md has the checks a change must pass and the rules the code depends on.

Licence

MIT, copyright Horizon (FTC 36596). You may use Zenith, change it and share it, including in your own team's code and in commercial work. Keep the copyright notice and the licence text with any copy you pass on. The full text is in LICENSE.

Credits

• The bundled fonts are Jost, IBM Plex Sans and JetBrains Mono, each under the SIL Open Font Licence 1.1.
• Robot following is Pedro Pathing v3, and commands are Ivy, Pedro Pathing's command library, or SolversLib.

NOTICE lists every third-party credit and licence.

Zenith is built by Horizon (FTC 36596).