Bhanu1776

Agent Migrate — AI skill for Claude Code

AI community

Move your whole AI coding-agent setup between harnesses in one command: Claude Code & Codex → pi, opencode, oh-my-pi, prime-agent, Hermes.

How to install Agent Migrate

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Bhanu1776/agent-migrate and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Migrate does

Move your whole AI coding-agent setup between harnesses in one command: Claude Code & Codex → pi, opencode, oh-my-pi, prime-agent, Hermes. Skills, MCP, memory, hooks, guards and chats. Dry-run first, zero dependencies.

Alternatives in AI

Pilotfish — Multi-model orchestration layer for Claude Code — the frontier model plans, cheaper models execute, verificati 688 ★
Memory Forge Rs — Stop resetting satisfying AI chats — edit the memory instead 474 ★
Cplt — Sandbox for AI coding agents 114 ★

README

agent-migrate

Move your coding-agent setup from one harness to another with one command.

claude-code ─┐                      ┌─→ pi · oh-my-pi · prime-agent
codex ───────┴─→  neutral bundle  ──┴─→ opencode · hermes

Readers turn a harness's config into a neutral bundle. Writers turn the bundle into another harness's files. Each new harness needs one adapter, not one converter for every pair.

Supported

Target	Default dir	Notes
`pi`	`~/.pi/agent`	Bridge extension for memory, hooks, and guards. Uses `pi-mcp-adapter` for MCP.
`oh-my-pi`	`~/.omp/agent`	omp already reads Claude and Codex setups, so the tool adds only what omp can't see. Import chats with `omp --from-claude` / `--from-codex`.
`prime-agent`	`~/.prime/agent`	Native MCP. The only tool is `ipython`, so guards and hooks see `!cmd` lines and string literals only.
`opencode`	`~/.config/opencode`	Guards become `permission.bash` rules. A plugin runs hooks and memory. Chats come in through `opencode import`.
`hermes`	`~/.hermes`	`SOUL.md`, `config.yaml` blocks, and native shell hooks and approvals. Secrets go in `.env`. Chats are not migrated.

What it moves

Part	From	To pi (other targets map to their own native equivalents)
Instructions	`CLAUDE.md`, `AGENTS.md`	A marked block in `AGENTS.md`. Your own text is kept and backed up.
Skills	user and plugin skills	Symlinks for skills pi can't already see. Skills that were "off" stay off.
Prompts	`commands/`, `prompts/`	`/name` commands. Plugin commands get the plugin name as a prefix.
MCP servers	`.claude.json`, `config.toml`	`mcp.json` (mode 0600) plus `pi-mcp-adapter`
Memory	Claude project memory, Codex memories	`memory/`. It goes into the prompt on every turn.
Hooks	settings, plugin hooks, Codex `notify`	A bridge exte