syncai

---

SyncAI is a lightweight and cross-platform utility that keeps AI-assistant guidelines, rules and ignored files in sync across multiple agents:

• Cursor
• GitHub Copilot
• Claude Code
• OpenAI Codex
• OpenCode

It watches the files you specify in a JSON configuration and propagates every change to the corresponding locations for the other agents.

Supported sync types

SyncAI supports four kinds of synced items in agent configurations:

• context — a single file with general AI guidelines or assistant context (for example AGENTS.md or CLAUDE.md). Configure with context.path. The file is copied verbatim to the target location.

• rules — a pattern that matches multiple rule files (for example Cursor rules or Copilot instructions). Configure with rules.pattern. If the pattern contains a * wildcard, the wildcard is replaced with the source file's base name when copying.

• ignore — a single file with instructions telling the assistant what to ignore (for example .copilotignore). Configure with ignore.path. The file is copied verbatim.

• skills — a directory pattern matching skill folders (for example .claude/skills/*). Each match is treated as a whole skill: every file inside the directory (including nested subfolders, scripts, data, etc.) is mirrored to the equivalent location at every other agent that has a skills pattern. The * wildcard captures the skill's folder name and is substituted into the target pattern. Skill files are copied verbatim. Deletions are propagated and now-empty target directories are cleaned up automatically down to the configured base directory, which is preserved.

These sections can be used together for each agent to keep context, rule files, ignore files, and skill folders in sync across different assistants.

Quick start

1. Download a suitable binary from the GitHub Releases.
2. Copy syncai.json to your project and adjust the configuration for your agents.
3. Launch the binary in the project dir ./syncai.

Configuration

Launching with arguments

1. Use ./syncai -config {path_to_syncai.json} to start SyncAI with your custom configuration file.
2. Use ./syncai -workdir {path_to_working_directory} to specify a different working directory.
3. Use ./syncai -no-watch to sync your files only once, without watching for changes (useful for CI).
4. Use ./syncai -self-update to update SyncAI to the latest version.

Configuration File

The default configuration is a simple JSON map (for more details check syncai.json):

{
  "config": {
    // sync interval in seconds
    "interval": 5,
    // working directory (optional, default is current directory)
    "workdir": ""
  },
  "agents": [
    {
      // agent name
      "name": "",
      // optional "rules" section
      // GitHub Copilot calls it "instructions", Cursor "rules"
      "rules": {
        "pattern": "./rules/*.md"
      },
      // optional "context" section
      "context": {
        "path": "/path/to/your/guidelines.md"
      },
      // optional "ignore" section
      "ignore": {
        "path": "/path/to/your/ignorefile"
      },
      // optional "skills" section: directory pattern matching skill folders.
      // The `*` is captured as the skill folder name and substituted on the
      // other agents. Every file inside the matched directory (recursively)
      // is mirrored.
      "skills": {
        "pattern": "./skills/*"
      }
    }
  ]
}

How it works

1. SyncAI loads the configuration file and builds a watch-list of directories and files derived from all sections.
2. It periodically scans those files for new or updated modification times.
3. When a rule file changes, its contents are copied to every other agent’s rule directory.

The copying logic is intentionally simple and conservative:

• The filename is preserved exactly, unless the target pattern contains a * wildcard—in that case, the wildcard is replaced with the source file’s base name.
• Destination directories are created as needed.

How to build

To build SyncAI manually, follow the next steps:

cd syncai

make build