# Laand MCP server

The full guide for people is at https://laand.me/docs/mcp. This is the same guide as plain text, for AI agents.

Use Laand from Claude, Claude Code, Codex, Cursor, ChatGPT or any MCP client. Hiring teams work with their Talent Hunts, candidates and submitted micro-projects (proof of work, written or code) where they already work.

Server URL:

```
https://laand.me/mcp
```

Sign-in is OAuth with your Laand company account. Each person connects their own agent, and it acts as them with their role in the workspace. No API keys.

You can also find these instructions in Laand under **Overview › Integrations › AI agents**, where you can see and disconnect your agents.

## Connect your client

**Claude (web or desktop):** Settings › Connectors › Add custom connector. Paste the URL, connect, then sign in to Laand and press Allow.

**Claude Code:**

```
claude mcp add --transport http laand https://laand.me/mcp
```

Then run `/mcp`, pick `laand` and sign in.

**Codex:**

```
codex mcp add laand --url https://laand.me/mcp
codex mcp login laand
```

**Cursor:** add to your MCP settings (`mcp.json`):

```json
{ "mcpServers": { "laand": { "url": "https://laand.me/mcp" } } }
```

**ChatGPT:** Settings › Apps & Connectors › Advanced, turn on developer mode, then create a connector with the URL and OAuth sign-in.

## Tools

| Tool | What it does |
|---|---|
| `list_hunts` | Your Talent Hunts and how many candidates are at each stage |
| `get_hunt` | One hunt: milestone, analysis, core problems, micro-projects and what each tests |
| `list_candidates` | Candidates in a hunt, optionally one stage |
| `get_candidate` | Scores, stage, stipend, the team's say, the work review's recommendation |
| `get_submission` | A submitted micro-project: the brief, Laand's review and, once the candidate is unlocked, the deliverable: the full write-up, or a `git clone` command for the code repo |
| `waiting_on_you` | Decisions your team is the blocker on |
| `ask_hiring_data` | Questions about your hiring, answered from the same numbers as Analytics |
| `leave_opinion` | Your Yes / Not sure / No on a candidate, with a note |
| `invite_to_micro_project` | Invite to an unpaid micro-project (owners and admins) |
| `move_to_final_stage` / `remove_from_final_stage` | Move a candidate in or out of the final stage (owners and admins) |
| `release_stipend` / `flag_stipend` | Release a stipend now, or stop it during its 48-hour hold (owners and admins) |

## Rules

- Candidates are anonymous (a made-up name) until your company has settled the hunt's fee **and** that candidate has submitted their micro-project. Your agent sees exactly what you'd see in Laand.
- A submission's deliverable is handed over once the candidate is unlocked. For code, the clone command gives read-only access to that one repo for an hour.
- Paying the fee, inviting to a paid micro-project and marking the hire happen in Laand. The tools answer with a link.
- Actions from an agent are recorded as yours "from an AI agent", in Laand and in your hunt's Slack channel.
- Members (not owners or admins) can read everything and leave their say, as in Laand.

## Example prompts

- "Which Backend candidates are waiting on me?"
- "Pull submitted micro-projects, clone the repo, run the tests and show me the weak spots."
- "Compare the top three submissions and draft interview questions."
- "Where do our best candidates come from?"