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

# Get started with AgentLinkOps

> Install the agent skill, connect your agent, verify the connection and check a supplied backlink. Four steps; the same ones the README and the landing page show.

Four steps and your agent can check a placement, read its dated evidence and keep watching it.
The steps are the same wherever you start: this page, the npm README, or the landing page.

<Note>
  Hosted monitoring needs a pilot invitation; the local check in step 4 needs no account.
  `agentlinkops` 0.5.0 is on npm and the skill discovery index is live on both domains, so both
  install lines below work as written. [Availability](/capability-status) tracks releases.
</Note>

## 1. Install the skill

```bash theme={null}
npx -y agentlinkops agent setup
```

This detects the agent clients on your machine (Claude Code, Codex, Cursor, Gemini CLI,
Hermes), installs the AgentLinkOps skill pack at each client's skills path and writes the MCP
entry for the view that client should use. It stores no credential and prints exactly what it
changed; `agentlinkops agent status` shows it again later.

If you use the `skills` CLI, `npx -y skills add https://agentlinkops.com` installs the same
skills; then add `https://app.agentlinkops.com/mcp` as a remote HTTP MCP server in your client
by hand ([MCP setup](/guides/connect-mcp)).

The skill is a short door. The first time your agent uses AgentLinkOps in a session it reads
the full [agent reference](/reference/agent-reference) (also `agentlinkops skill`), which
carries the rules, the observe-act-observe loop and the common mistakes.

## 2. Sign in when your agent asks

Your client opens the AgentLinkOps sign-in in the browser. Pick the account, the workspace and
the scopes for the task. The grant stays in the client's credential store; nothing is pasted
into a chat. A terminal or HTTP client uses a scoped API key instead ([CLI](/guides/connect-cli),
[HTTP](/guides/connect-http)).

## 3. Verify

Ask your agent:

```text theme={null}
List my AgentLinkOps projects.
```

A project list means the connection works. On the CLI, `agentlinkops doctor` reports the same
in one line per check. If nothing comes back, [MCP connection help](/guides/connect-mcp) covers
the three failure shapes.

## 4. Check a placement

Ask your agent, with real URLs:

```text theme={null}
Check whether [publisher page URL] links to [my target URL].
Show the completed observation, its date and any uncertainty,
and explain the check allowance before creating a watch.
```

The [placement workflow](/guides/check-placement) creates a monitored watch and requests a
check. A queued job is still waiting for a result; your agent waits for the completed
observation before reporting what the page contains.

No account yet? The same verifier runs locally:

```bash theme={null}
npx -y agentlinkops check --source https://publisher.example/resources --target https://your-site.example/guide
```

[Check one backlink locally](/guides/first-result) explains the result and how to keep it.

## Read the result

* `present`: the completed read found a matching link.
* `absent`: that read found no match. Check the URLs and scope before drawing a conclusion.
* `unknown`: the verifier could not decide. Read the reason and inspect the page when needed.

One absent observation does not establish confirmed loss. [Evidence states](/evidence)
explains how repeated observations affect that conclusion.

Keep the watch active to [monitor changes](/guides/monitor-changes), or pause it with
[update\_link\_watch](/reference/commands/update_link_watch). Pausing preserves its history.