Two paths. One command each.

BringYour maps every supported AI tool through a single intermediate format (a signed bundle), then emits whatever layout the target tool expects. Two real walkthroughs below — nothing hand-waved.

Path 1 — Migrate an existing harness

You already have an AI tool set up the way you like it and you want that same shape in a new tool. Run one command.

npx portable migrate --from claude-code --to cursor

Under the hood: BringYour reads the source tool's directory, scrubs PII (regex + blocklist + identity heuristics; every removal logged), builds a Bundle, signs it with ed25519, then writes whatever layout the target tool expects.

What gets read from Claude Code

~/.claude/
├── CLAUDE.md              # identity + preferences
├── agents/
│   └── code-reviewer.md   # frontmatter-prefixed subagent
└── skills/
    └── sample/
        └── SKILL.md       # frontmatter + body

What gets written into Cursor

project-root/
├── .cursorrules           # distilled top-level rule
└── .cursor/rules/
    ├── skill-sample.mdc
    └── agent-code-reviewer.mdc

Cursor uses .mdc files with front-matter keys Cursor understands (description, globs, alwaysApply). BringYour maps each Claude Code concept to the closest Cursor primitive — skills and agents become scoped rules.

Supported pairs

Bidirectional (read + write, round-trip tested): Claude Code, Cursor, Codex, OpenClaw, Aider, Continue, Cline, Goose, Zed, Roo Code. Export-only (paste into web UI): ChatGPT, Claude.ai. Write-only: Copilot.

---

Path 2 — Bootstrap a new harness from GitHub

You're new to AI tooling and staring at an empty CLAUDE.md. BringYour reads your public GitHub — repos, languages, topics, commit cadence — and drafts a first-pass harness that sounds like you.

npx portable bootstrap --github yourname

What BringYour reads (all public, unauthenticated)

• Your top-30 repos' primary languages (Go, Python, TS…)
• Repository topics and descriptions
• Commit cadence (avoids inferring "senior" from a 2-week account)
• README prose signals (tone, technical vocabulary)

What gets drafted (Claude Code target)

~/.claude/
├── CLAUDE.md                    # identity + preferences inferred from signals
├── memory/
│   └── user_profile.md          # "Go backend engineer, 7 yrs, ships
│                                #  CLIs, dislikes heavy abstractions"
└── skills/
    └── quickstart/SKILL.md      # one neutral starter skill

Nothing about the draft is permanent. You open it, fix what's wrong, move on. It beats a blank file.

---

Where the MCP server fits

The same two paths work from inside any MCP-aware agent (Claude Code, Cursor, Codex). Point the agent at our remote server and it can install or migrate without leaving the chat:

claude mcp add --transport http bringyour https://bringyour.ai/mcp

After that, the agent can call list_targets, install_harness, migrate_harness, or bootstrap_from_github directly. The MCP tools return file maps — the agent writes them using its own file-write tool, so nothing on our side ever touches your disk.

What the agent sees unauthenticated

• initialize — works
• tools/list — works, returns the 4-tool catalog
• tools/call — returns a license required error with the buy link

Discovery without payment, usage after. One $49 purchase unlocks all four tools for life.

Buy the $49 license →
← back to bringyour.ai