Contents
The short version
- Whatever an agent, key or token does under your account, you did.
- Give each connection the narrowest scope it needs, keep keys secret, and revoke what you no longer use.
- Agents can be wrong. Review what they change before it is published.
- Third-party AI tools are governed by their own terms, not ours.
A summary for convenience, not part of the document. The numbered sections below govern.
What these terms cover #
These terms are part of the Terms of Service. They cover every way the products are used by software: AI agents connected over the Model Context Protocol (MCP) to Terminal, the Hub and Baselumen; the in-app agent, which runs on an AI provider key you supply; content API keys and the content API; site-bound API keys; webhooks; and apps installed from the Marketplace that act on a workspace ("connections").
You are responsible for your connections #
- Anything done through a connection under your account or workspace is done by you, under the Terms of Service and the Acceptable Use Policy, whether you instructed it, reviewed it, or neither.
- Connecting an agent is an authorisation you give. Choose the scopes it gets, and grant only what the job needs. A connection may do anything its scopes allow.
- Keep keys and tokens secret. Do not put them in client-side code, public repositories or prompts shared with others. A content key is shown once; if one leaks, revoke it at once.
- Revoke connections you no longer use. Removing a member does not revoke a key they created for the workspace.
Agent output #
Agents, ours or anybody else's, can be wrong, incomplete or inconsistent, and can act on instructions hidden in content they read. Their output is not legal, financial or professional advice. Page and style changes made through a connection are saved as drafts until published; review them before you publish. CMS writes made with publishing switched on go live at once.
Stacklumen is not responsible for any action taken, content created or change published by an agent or connection, or for any loss that follows from one, to the fullest extent the law allows.
Third-party AI tools #
The AI clients you connect are made by third parties, and their terms and privacy policies govern what you send them and what they do with it. Data an agent reads from your workspace leaves Stacklumen when the agent reads it. The in-app agent sends what you ask it, and the workspace data its tools read, to the AI provider whose key you supplied. We do not supply, run or pay for the model behind any agent you connect or configure.
Limits and enforcement #
- Connections are subject to rate limits, scopes, plan limits and usage caps. Connecting an agent needs a paid plan once billing is switched on.
- We may throttle, suspend or revoke any connection, key or token at once, without notice, if it breaks these terms, threatens the service, or behaves abnormally.
- Prompt injection, attempts to extract system instructions or other customers' data, and use of a connection outside its scope are breaches of the Terms of Service.
Agents on engagements #
On an engineering engagement, Stacklumen staff may use agents to draft and make changes to a client's site or application. Those agents are directed by staff, and changes are reviewed by staff before they reach production. Clients do not direct them; to ask for a change, use your project's chat or write to support. Agent messages are not offers or commitments: scope, pricing and commitments bind only when confirmed by staff under the engagement's agreement.
These are the binding terms Stacklumen operates by. The short version at the top of each document is a summary and does not change it. Nothing here is legal advice to you; if a decision depends on these terms, ask your own counsel. Earlier versions are available on request, and anything that looks wrong or unclear can be raised at support@stacklumen.com.