Upwork introduced a new MCP Beta Participation Terms document governing access to its Model Context Protocol server, which allows third-party AI agents to connect to Upwork and perform limited actions on users' behalf. The beta enables agents to read information and prepare actions (such as searching jobs, drafting proposals, or sending messages) but restricts binding actions like contract creation or money movement to require direct user authentication on Upwork's platform. Users must accept these terms to participate in the invitation-only beta, and all agent-generated communications must carry a mandatory disclosure label.
Beta participants: Every message, proposal, job post, or invitation your AI agent sends to other Upwork users must be labeled to show it was AI-generated, and you cannot remove, obscure, or misrepresent that label.
Beta participants: You will be charged for every action your agent takes (such as submitting a proposal), even if you reject the draft before it is sent.
The updated terms establish a new framework for AI agent access to Upwork accounts during a limited beta period. Agents can perform read-only actions (searching jobs, viewing profiles) and draft-confirm actions (preparing proposals or messages for user approval) without per-action confirmation, but cannot execute binding actions like hiring or moving money—those require direct user authentication on Upwork's platform. All agent-generated communications visible to other users must carry a mandatory disclosure label. Users are responsible for all agent actions as if they took them directly, and all actions are charged to the user's account regardless of whether the user approves a draft-confirm action. You can revoke agent access through the App Authorization page.
→ Review the mandatory disclosure requirements for agent-generated communications before connecting an AI agent to your Upwork account.
→ Understand that you will be charged for all agent actions, including draft-confirm actions, regardless of whether you approve them before they are sent.
→ Revoke agent access at any time through the App Authorization page if you no longer wish to participate in the beta.
All agent-generated communications visible to other users must carry a clear, standard label disclosing agent involvement; removal or misrepresentation of the label is prohibited.
Agents cannot execute binding actions (contract creation, payment); these actions must complete on a logged-in Upwork page under direct user authentication.
Users are charged for all actions taken by the agent, including draft-confirm actions, regardless of whether the user approves the draft before it is sent.
This change record describes what was added, removed, or modified in the document. Analysis reflects what the updated agreement states or permits. It does not constitute a legal determination about enforceability. Applicability may vary by jurisdiction. Methodology
Wording edits to this record that do not change what it claims. Changes to a claim are published as corrections.
Upwork has introduced a new beta program framework for AI agent integration that creates disclosure and authorization obligations for beta participants. The terms establish that agent-generated user-visible communications must carry mandatory disclosure labels, and that …
Regulatory exposure, obligation change, escalation trigger, board-ready language, and recommended action for legal and compliance teams.
Unlock the full institutional analysis — InsightConductAtlas provides verified policy intelligence sourced directly from platform documents. All analysis is intended to support, not replace, legal and compliance review. Record CA-X-003368.
Upwork updated its Terms of Service and Escrow Instructions to clarify how service contracts are formed and to introduce new …
Upwork introduced a new beta program called 'Upwork Now' that enables faster, session-based hiring workflows for eligible clients and freelancers. …
Get the weekly coverage letter
Companies change their terms without announcing it. We capture governance documents, compare each new version with the last, and publish what changed, with the source text behind it. One email a week on what we found and what we could not reach.
Get alerted when this policy changes again, including what changed and why it matters.