Giving an AI agent access to your LinkedIn account is safest when the agent never holds your password or session at all, and works through a tool that runs inside your own logged-in browser. An agent that signs in as you from its own cloud computer runs straight into what LinkedIn's User Agreement says about sharing your account, however good its own safeguards are.

Reach is a LinkedIn tool that runs as a Chrome extension inside the session you are already signed into, so an AI agent can read your network, plan and draft over a fixed set of tools without ever being handed your LinkedIn login. This post looks at the question through the biggest agent launch of the last fortnight, and through our own account's numbers.

What did OpenAI launch with dots?

Always on. Their own browser.

On 29 September 2026 OpenAI introduced dots, always-on agents inside ChatGPT. In OpenAI's words they "have their own cloud computer, learn from feedback over time, and can work towards your goals 24/7", and through plugins "they can readily connect to over 4,000 apps." The launch page goes further: "Dots can do nearly anything using their own cloud computer, their own browser, and the apps you've connected."

They are rolling out to Pro, Business Premium and Enterprise plans. Pro users in the European Economic Area, Switzerland and the UK are excluded for now.

OpenAI has clearly thought about safety. When a dot is working in the background it uses tools "restricted to be read-only", so it cannot send messages there. Custom Rules let you allow, block or require approval for specific actions. An auto-review step checks actions "that could affect your accounts or share information". And for signing in, "dots can use saved passwords without exposing them to the model."

That last line is the one that matters for LinkedIn. It means the agent can sign in to a website as you, from a computer that is not yours.

Can an AI agent log into my LinkedIn account?

Technically, yes. Any agent with its own browser and your saved password can open LinkedIn and sign in. Whether it should is a different question, and LinkedIn has written its answer down.

LinkedIn's User Agreement, effective 3 November 2025, says in section 2.2: "You will not share your account with anyone else". In the list of things members agree not to do, it names using "another's account (such as sharing log-in credentials or copying cookies)".

LinkedIn has not said publicly whether an AI agent signing in on your behalf counts as "anyone else". We would not want to be the account that finds out. A LinkedIn profile carries every relationship and every conversation you have had there, and the agreement already describes handing the login to something else as the thing not to do.

Is it safe to give an AI agent access to your LinkedIn account?

It depends on what kind of access. Three designs are on the market and they carry very different risk.

The agent gets your login. A cloud agent, or a cloud automation service, signs in as you from its own machine. Your credentials or cookies now live somewhere you cannot see, and LinkedIn sees your account active from a computer you do not own.

The agent gets your browser. You let the agent drive your own computer. Dots supports this too: you can "give your dot permission to connect and use your laptop". The login stays with you, but the agent can now do anything your browser can do, on every site you are signed into.

The agent gets a tool list. The agent never touches LinkedIn. It talks to a tool that already sits inside your own session, and that tool decides what the agent may ask for. This is the design we built Reach on, and the one we would recommend checking for in anything you connect.

Being a bit honest with ourselves about the third option. The same User Agreement also prohibits using "bots or other unauthorized automated methods" to "send or redirect messages". That sentence covers browser extensions as well, ours included. Running in your own browser does not exempt anyone from it, and no tool can make a restriction impossible. What the design changes is who holds the keys, and how much damage a mistake can do.

How does Reach give an agent LinkedIn access without your password?

The work is split in two. The agent and the Reach server hold the intent: who to contact, what to say, when, and whether a human has approved it. The Chrome extension in your own browser holds the hands. It is the only part of Reach that ever talks to LinkedIn.

In practice the extension checks for approved work about every 60 seconds, takes one job at a time, does it inside the session you are already signed into, and reports back. The Reach server never calls LinkedIn, and it stopped storing session cookies altogether on 14 September 2026. If your laptop is shut, queued work simply waits. Nothing takes over from a server.

The agent, whether Claude, ChatGPT or another assistant, connects over MCP and gets a fixed vocabulary of tools. It can read the plan and the pipeline, qualify people, propose a message and read results. It cannot see your password, and it cannot raise its own limits. We walked through that tool list in detail in AI agent for LinkedIn: what should it be allowed to do?, and through the extension's permissions in Is a LinkedIn Chrome extension safe?

One rule cannot be switched off by anyone, including us: Reach will not contact a person it has already contacted. Five sends on our own account stopped mid-flight recently and Reach cannot prove whether they arrived. Nothing was resent. They sit parked until a person opens the thread and looks.

Why does an always-on agent need the network you already have?

Because the bottleneck on most LinkedIn accounts is follow-through, and more hours do not fix it. Our own account has 16,336 people in its network and 2,351 of them have replied to us at some point. The inbox holds 3,374 threads, and in 2,069 of them the last message is theirs, not ours.

These are people who already know us. Working them is reading, ranking and drafting, which is what agents are good at. The sending is the part to keep close.

The honest counterweight from the same account: in the last seven days 39 messages and 55 invites went out from this profile, and by Reach's own estimate about 61 of those 94 actions were done by hand on LinkedIn. Reach sent 4 messages and 29 invites. The agent does the reading and the first draft. A human still decides what goes.

Common questions

Should I give ChatGPT or another AI agent my LinkedIn password?

No. LinkedIn's User Agreement tells members not to share their account and lists sharing log-in credentials as a way of using another's account. Use a tool that keeps the session in your own browser and gives the agent tools instead of a login.

Can OpenAI's dots use LinkedIn?

A dot has its own browser and can sign in to supported websites with saved passwords, so in principle it can open any site you give it access to. Whether that is acceptable for LinkedIn is governed by LinkedIn's terms, and those terms say not to share your account.

Does Reach need my LinkedIn password?

No. Reach runs as a Chrome extension inside the LinkedIn session you are already signed into. The Reach server never calls LinkedIn and holds no LinkedIn session cookies.

Can an AI agent get my LinkedIn account restricted?

Yes. LinkedIn's terms prohibit unauthorized automated sending whoever is driving, and an agent that sends like a bot looks like one. Keeping the work in your own browser, going slowly, with a person approving each message lowers the risk. Nothing removes it, and any vendor claiming otherwise is selling you something.

What happens to Reach's work when my browser is closed?

It waits. The extension is the only worker, so nothing runs on LinkedIn while your browser is closed, and queued work picks up when you are back.

Reach is at reach.linkenite.com.


Sources

Every URL below was opened and read on 2026-10-08. Quotes are verbatim from the opened page.

Trend

  • OpenAI, "Introducing dots", 29 September 2026 - https://openai.com/index/introducing-dots/ (read in full in-browser)
  • "Powered by GPT‑6 Astra, they have their own cloud computer, learn from feedback over time, and can work towards your goals 24/7."
  • "they can readily connect to over 4,000 apps"
  • "Dots can do nearly anything using their own cloud computer, their own browser, and the apps you've connected."
  • "you can also give your dot permission to connect and use your laptop"
  • "For signing into supported websites, dots can use saved passwords without exposing them to the model."
  • Proactive research uses tools "restricted to be read-only, which means that they can't send messages, change app content, or control your browser or computer."
  • Custom Rules "let you allow specific actions, require approval, or block them"; auto-review checks actions "that could affect your accounts or share information"; "Certain sensitive tasks, such as changing a password, always stay with you."
  • Availability: rolling out to Pro, Business Premium and Enterprise; Pro excluded in the EEA, Switzerland and the UK. Corroborated by The Next Web, 29 Sep 2026 (https://thenextweb.com/news/openai-dots-always-on-ai-agents-cloud-computers-devday) and MacRumors, 29 Sep 2026 (https://www.macrumors.com/2026/09/29/openai-launches-dots/).

Primary

  • LinkedIn User Agreement, effective 3 November 2025 - https://www.linkedin.com/legal/user-agreement (read in-browser, text extracted verbatim)
  • Section 2.2: "You will not share your account with anyone else and will follow our policies and the law."
  • "not share or transfer your account or any part of it"
  • Don'ts: "use or attempt to use another's account (such as sharing log-in credentials or copying cookies)"
  • Don'ts: "Use bots or other unauthorized automated methods to access the Services, add or download contacts, send or redirect messages, create, comment on, like, share, or re-share posts, or otherwise drive inauthentic engagement"

First-party (Reach MCP, account pravin@linkenite.com, 2026-10-08 ~02:30 UTC)

  • get_account_state: pipeline total 16,336 people, 2,351 replied; last 7 days messages 39 (4 by Reach, 35 manual), invites 55 (29 by Reach, 26 manual) - "manual is inferred as total minus what Reach can prove it sent, so it is an estimate"; 5 uncertain sends parked, nothing resent; extension worker online in the owner's browser.
  • get_inbox: totalThreads 3,374, totalNeedingReply 2,069 (threads where their message came after ours).
  • get_gates: duplicate-protection "Already contacted this person" is always on and cannot be switched off.

Product facts

  • Reach docs/STATE-OF-THE-SYSTEM.md on master (7965d41): "Reach never touches LinkedIn from a server"; extension "polls /api/extension/job/next every 60 seconds, leases one job"; server "holds no LinkedIn credentials: it stopped storing or returning session cookies on 2026-09-14".

Internal links (all returned 200 on 2026-10-08)

Other Blogs

2024-03-26
Operational Optimization
AI-Driven Insights for Modern HR Management

Artificial Intelligence (AI) is transforming industries, and Human Resource Management (HRM) is no exception. But how exactly is AI reshaping HR practices? Let’s delve into the key trends, benefits, and future directions of AI in HRM.

Read More
2024-03-26
Human-in-the-Loop Workflows
The Importance of Human Intervention in AI-Driven Workflows

LLMs are designed to predict the next word or sequence based on vast amounts of training data. This predictive capability, while powerful, is inherently prone to errors

Read More
2024-03-26
Human-in-the-Loop Workflows
Enhancing the Reliability of GPT-Assisted Market Research through Human-in-the-Loop Methodologies

The rapid advancements in artificial intelligence, particularly with Large Language Models (LLMs) like GPT (Generative Pre-trained Transformer), have revolutionized market research.

Read More
2024-03-26
Operational Optimization
Leveraging Human-in-the-Loop AI for Reliable Supply Chain Innovation

The emergence of generative AI tools like ChatGPT has sparked tremendous excitement and opened up a world of possibilities for how businesses operate. While the potential applications for AI in the supply chain are

Read More
2024-03-26
Human-in-the-Loop Workflows
Sales Enablement with Human-in-the-Loop AI

In today's fast-paced business environment, advancements in artificial intelligence (AI) have significantly transformed the sales landscape.

Read More
2024-03-26
AI Strategy and Consultation
How Human-in-the-Loop AI Enables Customer Engagement and Marketing

In the fast-paced world of digital marketing, businesses are constantly seeking innovative ways to engage with their customers and stay ahead of the competition. Generative AI, such as GPT, has emerged as a powerful tool

Read More
2024-06-06
AI-Powered Solutions
Strategic Approaches to Leveraging AI Innovations

2024 brings transformative trends that will shape the future of technology and business. From multimodal AI to ethical AI development, understanding these trends is crucial for staying competitive. Discover how open-source frameworks are democratizing AI, how customization enhances user experiences, and why edge AI is revolutionizing data processing. 🚀 To dive deeper into these insights and strategic approaches, click on "Read more" below: Key Takeaways: Multimodal AI: Integrates text, image, and audio data for improved accuracy. Open Source AI: Accelerates innovation and reduces costs. Customization: Tailors AI solutions to specific needs for better outcomes. Edge AI: Enhances performance and privacy in real-time applications. AI in Cybersecurity: Protects against sophisticated threats. Ethical AI: Ensures transparency, fairness, and compliance. Stay ahead of the curve by leveraging these AI and machine learning trends in 2024. Embrace the future of technology and drive innovation in your business! 💼💡

Read More
Quick Contact