agentsync

Define your MCP servers, memory, skills, and plugins once. Run agentsync apply. They land — correctly translated — in every AI coding agent on your machine.

Get started Your first sync View on GitHub

If you use more than one AI coding agent, you keep re-entering the same config in each one’s native format, you forget to install a plugin everywhere, and your configs quietly drift apart. agentsync fixes the fan-out.

Edit once, apply everywhere

Add an MCP server or install a plugin a single time. It’s projected into each agent’s native config — fully where the agent has the concept, lossily but reported where it doesn’t, skipped (never silently) where it can’t.

Bidirectional, chezmoi-style

When an agent edits its own config, agentsync detects the drift and offers a merge: adopt the edit into your source, or re-impose the source. Nothing is overwritten behind your back.

Secrets stay secret

Reference ${secret:github.token}. agentsync resolves it at apply time from an age-encrypted vault and never writes the cleartext back into your (committable) source.

Offline & reproducible

apply never touches the network — it renders from a local cache. Only update reaches out. Fast, deterministic, CI-friendly.

From zero to synced in five commands

agentsync init                          # create ~/.agentsync/
agentsync agent add claude              # register the agents you use
agentsync agent add opencode
agentsync mcp add github \              # add a server once…
  --command npx --args "-y,@modelcontextprotocol/server-github"
agentsync apply                         # …it fans out to every agent

That github MCP server now lives in both ~/.claude.json and ~/.config/opencode/opencode.json, each in its own native shape.

Find your path

What is agentsync? The problem it solves and the promise it makes — in two minutes.

Your first sync The 5-minute greenfield path: start clean, end with a server live in two agents.

Already have configs? Bring your existing Claude / OpenCode setup under management without retyping it.

The mental model Three states, one comparison. Read this once and everything clicks.

Secrets Reference credentials safely with an age-encrypted vault that never leaks cleartext.

Capability matrix Exactly what each agent supports, per component — and what's lossy or deferred.

Supported agents

Agent	Status	Component coverage
Claude Code	✓ full adapter	All seven components, including LSP.
OpenCode	✓ adapter	MCP, memory, skills, subagents, commands. Hooks + LSP skipped.
Codex CLI	🔜 planned	No-op today; `agent add codex` rejected unless overridden.
Cursor	🔜 planned	No-op today; will project skills, subagents + project-scope rules.