# Onboard yourself to Vamo

You are an AI agent. Work through the steps below in order, then report back to
the user. Do not skip step 2: it proves the connection before anyone hands you a
credential.

Vamo is a recruiting platform for engineering talent. Through this server you can
Search GitHub for developers, read their public work, build shortlists on a
project, and run outreach sequences.

## 1. Add the server to your client

- Endpoint: `https://mcp.vamotalent.ai/mcp/platform`
- Transport: remote MCP over Streamable HTTP
- Auth: an `Authorization: Bearer <key>` header

You probably do not have a key yet. Add the server now with the placeholder
below and replace it in step 3. Documentation tools answer without a key, so the
first working call costs the user nothing.

### Claude Code

```sh
claude mcp add vamo https://mcp.vamotalent.ai/mcp/platform --transport http --header "Authorization: Bearer VAMO_API_KEY"
```

### Cursor (`.cursor/mcp.json`), VS Code, Windsurf, Zed

```json
{
  "mcpServers": {
    "vamo": {
      "type": "http",
      "url": "https://mcp.vamotalent.ai/mcp/platform",
      "headers": { "Authorization": "Bearer VAMO_API_KEY" }
    }
  }
}
```

### Claude Desktop (`claude_desktop_config.json`)

Claude Desktop reaches remote servers through `mcp-remote`:

```json
{
  "mcpServers": {
    "vamo": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.vamotalent.ai/mcp/platform",
        "--header",
        "Authorization: Bearer VAMO_API_KEY"
      ]
    }
  }
}
```

Any client that speaks remote MCP over Streamable HTTP works with the same URL
and the same header.

## 2. Confirm the connection with a documentation call

Reload your MCP servers, then call `explore_docs`. Documentation tools need no
authentication, so a successful response here means the transport, the URL, and
the header plumbing are all correct, and the user has spent nothing.

If the call fails, fix the connection before going further. Report the exact
error to the user rather than guessing at a key.

## 3. Ask the user for an API key

Say this to the user:

> Create an API key at https://app-staging.vamotalent.ai/settings/api-keys, grant it the permissions you
> want me to have, and paste it here.

Then put the key in the client config from step 1, replacing `VAMO_API_KEY`, and
reload the server. Store it where your client stores credentials. Do not write it
into a file that gets committed.

If the user asks what to grant, tell them: read on Search GitHub and projects is
enough to source candidates; write on outreach is needed before you can send
anything.

## 4. Verify with one authenticated call

Call `account_execute` and read the account back. That is a cheap read and it
confirms the key resolves and carries permissions.

## 5. Report what you can now do

Tell the user which topics answered, what the key's permissions allow, and what
you would need granted to do more.

## How these tools work

There are four topics: Explore, Outreach, Projects, and Account. Each one has a
pair of tools:

- `<topic>_docs` returns typed documentation for the endpoints in that topic.
- `<topic>_execute` runs JavaScript you write in an isolated sandbox, with a
  client already bound to the user's account.

So you are not calling hundreds of endpoints one at a time. You read the
documentation for a topic, then write one piece of code that chains the calls,
filters, and aggregates, and you get one result back. Always read a topic's
documentation before writing code against it. The documentation is generated from
the live API, so it is accurate at the moment you read it.

## Rules

- Documentation tools are free and unauthenticated. Execute tools act on the
  user's account and spend credits, priced exactly like the API.
- The key carries only the permissions the user granted. A call outside them
  returns 403. Do not retry it. Tell the user which permission is missing.
- Anything that contacts a real person goes through outreach and spends credits.
  Confirm with the user before you send.