Save a working AI setup once. Restore it on a new machine. Reuse it across tools.
Import · Bootstrap · Install · Why · Support · Docs
The AI workflow that "works" inside a team is usually real, valuable, and trapped on one laptop.
MyAI turns that fragile setup into a reusable local-firstprofile.
Alpha status: narrow by design. Shipping path today is Claude Code -> MyAI repo -> Codex CLI.
When a team says "our AI workflow works," what that often means is:
- one person figured out the setup
- everyone else copies fragments by hand
- a new machine means rebuilding from memory
- a new tool means starting over
MyAI exists to make the setup itself portable, not just the output.
- import a working Claude Code setup into a reusable profile
- search and list saved profiles locally
- apply a saved profile and roll it back safely
- bootstrap a new teammate or new machine from team defaults
- sync the supported Codex subset without rewriting config by hand
| Most AI repos save | MyAI saves |
|---|---|
| isolated prompts | a portable working profile |
| machine-local tweaks | materialized, rollback-safe local state |
| tool-specific configs | reusable assets plus explicit sync |
| tribal knowledge | searchable reuse and restore telemetry |
This alpha does not try to support every tool. It tries to prove that one valuable setup can survive tool switching, teammate onboarding, and machine resets.
| Good fit | Not the first target |
|---|---|
| engineering teams already using Claude Code and Codex | solo prompt collectors with no repeat workflow |
| teams onboarding new machines or teammates often | teams looking for hosted analytics first |
| operators who want files, Git, and local control | teams needing every AI tool supported on day one |
| Capability | Status | Notes |
|---|---|---|
| Import from Claude Code | yes | Reads CLAUDE.md and supported MCP config from ~/.claude.json |
| Local profile search/list/show | yes | Search telemetry is written to local event logs |
| Apply and rollback | yes | Materializes to .myai-applied/ with backup and restore flow |
| Sync to Codex | yes | Current target is the supported Codex MCP subset |
| Pilot reporting | yes | Repo-local summary for reuse, search, actor, and machine metrics |
| Cursor target | later, only after the current wedge proves out | Not implemented in v0.1 |
| ChatGPT / hosted sync | not yet | Deliberately out of v0.1 scope |
- Local-first — your data lives in
~/.myai/, a plain Git repo you own - Portable by default — stored data remains readable without MyAI
- Team-oriented — workflows are inheritable, not trapped with one person
- Narrow first — solve one portability path well before expanding
- Standard formats — Markdown + YAML, no vendor lock-in
Current release track: 0.1.0-alpha.
npm install -g @myai/cli
myai helpOr run without a global install:
npx @myai/cli helpRequirements:
- Node.js 18+
cd /path/to/your/claude-code-project
myai init
myai profile import --from claude-code
myai profile listIf you want a stable name instead of using the current directory:
myai profile import code-review --from claude-codeCLAUDE.md- matching project
mcpServersfrom~/.claude.json - it does not import
.claude/settings.local.json, local permissions, or chat history
git clone <your-team-myai-repo> ~/.myai
npm install -g @myai/cli
cd /path/to/your/work-repo
myai bootstrap team-default --target-dir . --yesIf the profile declares sync.targets: [codex], bootstrap also restores the supported Codex config.
myai profile list --repo ./examples/sample-repo
myai profile show code-review --repo ./examples/sample-repo
myai profile apply code-review --repo ./examples/sample-repo --target-dir /tmp/myai-demo --yesYou do not need any telemetry env vars for normal use. If you are running an internal pilot and want stable attribution across operators or machines, set:
export MYAI_ACTOR_ID=pilot-operator-1
export MYAI_MACHINE_ID=macbook-air-01myai report summary then surfaces unique actors, unique machines, search volume, zero-result searches, and search-to-reuse within 24 hours. Logs stay local under ~/.myai/logs/.
~/.myai/
├── prompts/ # Saved prompts (Markdown)
├── mcps/ # MCP server configs (normalized format)
├── preferences/ # Team and personal rules
├── profiles/
│ ├── team/
│ └── personal/
├── skills/ # Optional workflow templates (light support)
├── logs/ # Local event logs for pilot validation
└── myai.yaml # Repo config
It's just files. Back it up with Git. Share it with your team. Move it to a new machine in 30 seconds.
v0.1 focuses on one narrow promise:
A team can save a working AI profile in one place and restore or reuse it across people and machines without starting from zero.
- Primary portability path: Claude Code → Codex CLI
- Product surface: local repo + CLI
- Target users: AI-native engineering teams (5-30 members) using multiple AI tools
| Doc | What it covers |
|---|---|
| STRATEGY.md | Product strategy, ICP, wedge, pricing, GTM, kill criteria |
| PRD-v0.1.md | v0.1 scope, requirements, use cases, success metrics |
| repository-schema.md | On-disk repository layout, namespaces, logs, bootstrap defaults |
| profile-schema.md | Main product object schema and validation rules |
| cli-command-spec.md | CLI command contract, flags, defaults, and exit codes |
| claude-code-to-codex-mapping.md | Supported portability subset and explicit warning policy |
| fresh-machine-pilot-runbook.md | Step-by-step fresh-machine and pilot validation flow |
| internal-dry-run-2026-03-30.md | Repo-local dry run record and current validation gaps |
| github-launch-kit.md | Suggested GitHub repo description, topics, tagline, and social preview metadata |
| examples/README.md | Sample repository for demos, pilots, and manual testing |
| IDEA-archive.md | Early vision exploration (historical) |
MIT