> ## Documentation Index
> Fetch the complete documentation index at: https://docs.proofable.me/llms.txt
> Use this file to discover all available pages before exploring further.

# Import an agent

> Bring a package, repo, or card. Proofable wraps identity and connections.

Drop a package, pick a GitHub repo, or paste a card or runtime link. Proofable inspects the source, registers identity on your profile, and asks you to reconnect apps. It does not copy secrets or run the package.

Supported inputs include OpenClaw workspace files, Hermes config and skills, Agent Skills `SKILL.md` folders, A2A Agent Cards, MCP URLs, GitHub repositories, and zip or tar packages.

## From the product

1. Open **Import your agent**.
2. Drop files or paste a link. Do not start with a name.
3. Review the detected label, skills, and apps that need reconnect.
4. Continue to register identity. Set spend and action limits as a separate step.
5. Connect the runtime from the project:

```bash theme={"system"}
npx -y @proofable/sdk mount <agentId> --apply <host>
```

Hermes, OpenClaw, OpenCode, Cursor, Claude Code, and Codex imports use those `--apply` values. VS Code uses `--apply cursor`. The browser never writes host files.

## From the CLI

```bash theme={"system"}
npx -y @proofable/sdk auth --oauth
npx -y @proofable/sdk import ./agent-pack.zip
```

`import` accepts a path, a directory, or a URL. It prints a hosted register link. Secrets in `.env` files, keys, and token stores stay out of the preview.

## HTTP

Inspect first, then map onto identity create. Retry is safe. There is no separate MCP import tool.

```bash theme={"system"}
POST https://api.proofable.me/api/v1/agents/import/preview
POST https://api.proofable.me/api/v1/agents/import/commit
```

Send JSON: `url`, `text`, `filename`, `archiveBase64`, or `files`. Commit returns `identityPrefill` and `mcpCreateArgs` for [`proofable_agent_create`](/mcp/agent-create).

If a Proofable identity is already on the source, Proofable links it only after this profile owns it. Otherwise it registers a new agent and leaves the source unchanged.

## What stays out

Credentials, OAuth databases, private keys, and token values are excluded. Reconnect those apps with OAuth or an existing account connection. Spend limits stay an explicit review.

## Related

* [Agent identity](./agent-identity)
* [Connect agent context](./runtime-mount)
* [Agent create](../mcp/agent-create)
* [Proofable CLI](../sdks/cli)