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

# AI Assistants (MCP)

> Let ChatGPT, Claude, or Langdock answer questions about your candidates, jobs, and pipeline using your TalentUnveiled data.

Connect an AI assistant to TalentUnveiled and you can ask it questions about your hiring in plain language — *"Who are my top three candidates for the Engineering Lead role?"*, *"Which applications have been waiting more than a week?"*, *"Give me a side-by-side of these two candidates"* — and get answers backed by your real data, not guesses.

This works through the **Model Context Protocol (MCP)**, an open standard for connecting AI assistants to outside systems. You don't need to install anything on the TalentUnveiled side — just point your assistant at our server and sign in.

<Info>
  **What stays private.** You sign in with your normal TalentUnveiled account, and the assistant can only access your organization's hiring data — nothing from any other organization. See [Privacy and security](#privacy-and-security) below.
</Info>

## What you can ask

Once connected, your assistant can:

* **Find candidates.** *"Show me everyone who applied for the Marketing Manager role in the last week"* or *"search for candidates with the email domain @acme.com"*.
* **Review profiles.** *"Give me a summary of Sarah Chen's profile, including her interview highlights and CV scores."*
* **Read interview transcripts.** *"Show me the transcript of Sarah's interview for the Product Designer role."*
* **Dig into CV evaluations.** *"How does Alex's CV score against our criteria?"*
* **Compare candidates side-by-side.** *"Compare Alex and Priya for the Senior Engineer role."*
* **Check your pipeline.** *"What does the pipeline look like for the Product Designer role?"* or *"Which jobs have candidates waiting on interview review?"*
* **See what needs your attention.** *"What needs my attention today?"* — pending reviews, candidates waiting too long, and new applications in one answer.
* **Check a job's interview setup.** *"Which interview questions and evaluation criteria are configured for the Sales Manager role?"*
* **Move candidates through stages.** *"Move Sarah Chen to 'Shortlisted' for the Product Designer role."* (Most assistants will ask for confirmation before taking any action.)
* **See organization metrics.** *"How many active jobs do we have? How many interviews this month?"*

The assistant reads scores on TalentUnveiled's 1–5 scale and respects your evaluation criteria. A score of `null` means *not scored yet* (for example, the interview hasn't happened) — not a low score.

## Connect your assistant

Every client uses the same server URL:

```
https://api.talentunveiled.com/mcp
```

You'll sign in with your normal TalentUnveiled account the first time you connect — no API keys to create, no extra setup. Pick the assistant your team uses — you only need one. If you're free to choose, Claude is the quickest to set up.

<Note>
  **Using Microsoft Copilot?** TalentUnveiled can also be connected to Microsoft 365 Copilot as a read-only connector, but that setup is done by your IT administrator rather than from your own settings. [Contact us](mailto:support@talentunveiled.com) and we'll help your IT team get it running.
</Note>

### ChatGPT

Custom connectors in ChatGPT require **Developer mode**, available on the Plus, Pro, Business, Enterprise, and Edu plans in ChatGPT on the web (on a Business or Enterprise workspace, an admin may need to enable it first). Developer mode also lets the assistant *move* candidates between pipeline stages — ChatGPT asks you to confirm each change before it happens.

1. Open **Settings → Security and login** and turn on **Developer mode**
2. Go to **Settings → Connectors → Create**
3. Server name: `TalentUnveiled`
4. Server URL: `https://api.talentunveiled.com/mcp`
5. Authentication: **OAuth**
6. Click **Save**, then sign in through the browser tab that opens

Mention `@TalentUnveiled` at the start of a chat to bring the tools in.

<Note>
  Developer mode lets connectors do more than read — only turn it on for connectors you trust. TalentUnveiled is safe to use with it. See [OpenAI's documentation](https://developers.openai.com/api/docs/guides/developer-mode) if a setting has moved.
</Note>

### Claude

The quickest setup for most HR teams. Custom connectors are available on every Claude plan (the Free plan is limited to one), in Claude on the web and in Claude Desktop.

1. In Claude, open **Settings → Connectors → Add custom connector**
2. Name: `TalentUnveiled`
3. Remote MCP server URL: `https://api.talentunveiled.com/mcp`
4. Click **Save**, then click **Connect** — a browser tab opens for you to sign in to TalentUnveiled
5. After you sign in, TalentUnveiled appears in your connectors list. Open a new chat and the tools are ready to use.

<Note>
  On Claude **Team** and **Enterprise** plans, an owner can add the connector once for the whole organization under **Organization settings → Connectors** — each member then only clicks **Connect** and signs in.
</Note>

### Langdock

Best for teams who already use Langdock as their shared AI workspace.

1. In your Langdock workspace, go to **Tools → MCP Servers → Add server**
2. Server URL: `https://api.talentunveiled.com/mcp`
3. Transport: **Streamable HTTP**
4. Authentication: **OAuth Authentication**
5. Click **+ Add connection** and sign in with your TalentUnveiled account
6. Click **Test connection**, then pick which tools you want exposed and click **Save tools**

Everyone in the workspace can now use these tools from their assistants.

### Other MCP tools (for technical team members)

Any tool that supports MCP can connect, and the pattern is always the same: add `https://api.talentunveiled.com/mcp` as a remote MCP server (transport: streamable HTTP), then sign in through the browser tab that opens — same account, no API key. For example, in Anthropic's terminal-based [Claude Code](https://claude.com/claude-code):

```bash theme={null}
claude mcp add --transport http talentunveiled https://api.talentunveiled.com/mcp
```

The first call in a session opens a browser tab for sign-in; run `/mcp` inside Claude Code to confirm the connection. The same pattern works in Cursor, VS Code, Codex, Zed, and other MCP-compatible tools — check your tool's documentation for where to add an MCP server.

## Privacy and security

* **Your account, your organization.** You sign in with your normal TalentUnveiled account, and the assistant can only access your organization's hiring data. Every member of your organization who connects gets the same access.
* **Changes need your confirmation.** The only change an assistant can make is moving a candidate to a new pipeline stage — and assistants ask you to confirm before doing it. Everything else is read-only.
* **Your AI provider processes the data.** Your questions and the hiring data the assistant reads flow through your AI provider under your contract with them. Because this includes candidate data, we recommend a team or business AI plan, where providers exclude your data from model training by default.
* **Disconnect anytime.** Remove the connector in your assistant's settings and the assistant loses access.

## Troubleshooting

<Accordion title="The sign-in page won't open">
  Make sure your computer has a default browser set. On a managed laptop, your company's single sign-on may need you to be signed in to TalentUnveiled in a normal browser first — open the [HR portal](https://hr.talentunveiled.com), sign in once, then retry the connector.
</Accordion>

<Accordion title="The assistant says it can't find any jobs or candidates">
  Your account is signed in but not yet linked to your TalentUnveiled organization. Sign in to the [HR portal](https://hr.talentunveiled.com) once with the same account — that links you to the right organization — then retry. If it still doesn't work, contact your TalentUnveiled admin or [support@talentunveiled.com](mailto:support@talentunveiled.com).
</Accordion>

<Accordion title="The assistant returns 'null' for a candidate's score">
  This means scoring hasn't finished yet — usually because the interview hasn't taken place or the CV evaluation is still running. It's not a low score; there's just nothing to show yet. Try again once the interview completes.
</Accordion>

<Accordion title="I was asked to sign in again after a few days">
  Sign-in tokens expire after a period of inactivity for security. Sign in again when prompted — your settings and selected tools stay the same.
</Accordion>

<Accordion title="Can I use an API key instead?">
  No. The MCP server uses browser sign-in only. The API keys you can create under **Settings → Developer → API Keys** are for the separate REST API and won't work with MCP.
</Accordion>

## What's Next

<CardGroup cols={2}>
  <Card title="Application Lifecycle" icon="arrows-spin" href="/concepts/application-lifecycle">
    Understand how candidates move through the hiring pipeline.
  </Card>

  <Card title="Managing Applications" icon="file-lines" href="/guides/managing-applications">
    The same actions, done from the HR dashboard.
  </Card>
</CardGroup>