Access token The credential an agent holds after a successful connection. Renewed automatically; you never handle it.
Agent Any AI that creates on your behalf and connects to AgentGrid.io with its own scoped, revocable identity. Always acts for one human, and never uses a team seat. How it connects depends on what it can do: see chat assistant and self-connecting agent.
Anima CLI
The command-line tool coding and autonomous agents connect with. One command, npx @animaapp/cli@latest login, and nothing to install. See The Anima CLI.
Anonymous artifact Something made by an AI that never connected: live, read-only, owned by nobody, and expiring 24 hours after creation unless someone takes ownership of it.
Artifact The unit of work: a live app, document, or stored file backed by a git repository, with one address and one history.
Artifact type
app, markdown, or an asset. Decides how a human sees the artifact.
Asset Any artifact that is neither an app nor markdown: stored, versioned, shareable, readable by agents, with no preview. The hub's third type filter.
Capability
One of read, write, share, publish. Granted to an agent by a human, never exceeding what that human holds. Fixed at invite time: to change them, remove the agent and invite it again. Visible per agent under Share → Manage access.
Chat assistant The ordinary chat in ChatGPT or Claude, and anything running in a browser tab. Connects through the official AgentGrid.io connector (in Claude) or plug-in (in ChatGPT), installed from that assistant's own directory. Anything in a browser is a chat assistant, whatever its name. See How agents connect.
Claim Taking ownership of an anonymous artifact. Moves it into your team, unlocks editing, and stops the expiry. Applies when the AI that made it was never connected; a connected AI creates artifacts in your team directly.
Connector What the official AgentGrid.io integration is called in Claude. Installed from Claude's directory, with signing in part of the install. ChatGPT calls the same thing a plug-in.
Hermes A personal AI agent, and a self-connecting one. It connects with the Anima CLI login command.
Invite (agent) A one-time connection link minted from Share → Invite an agent, carrying a name and the capabilities you chose. Creating it connects nothing: you copy the link, or the ready-made prompt, and give it to the agent.
Invite (person) An email invitation to join your team at a role, sent from Share → Invite people to team.
Manage access The Share dialog pane listing everyone with access to the team: people with their roles, agents with their capabilities, and removal for both.
OpenClaw An open personal AI agent, and a self-connecting one. It connects with the Anima CLI login command.
Plug-in What the official AgentGrid.io integration is called in ChatGPT. Same idea as Claude's connector: installed from ChatGPT's connector directory, with signing in part of the install.
Publish Deploy an artifact to a public URL that anyone can open with no account. Separate from sharing, and always explicit.
Restore Making an earlier version the current one, in place, from the History panel. The history keeps everything, including the restore itself.
Revoke Cut off an agent's access, from Share → Manage access. Takes effect on its next request, not at a future expiry. Also the only way to change what an agent can do: revoke, then invite it again.
Role What a person holds in a team: Viewer, Editor, Admin, or Owner. Distinct from an agent's capabilities, and unlike them, changeable at any time.
Save The button in the document editor. Auto-save is not live yet, so a change is only stored once you select it.
Self-connecting agent
An AI that can connect itself: Claude Code, Codex, Cursor, Claude Cowork, OpenClaw, Hermes, worker packs, and desktop agents. One command connects it, npx @animaapp/cli@latest login. Nothing to install.
Share Give someone access: the artifact link for team members, a team invite, or an agent invite. Not the same as publishing.
Team The container artifacts belong to. You may belong to several and switch between them; the switcher shows your role and the team's plan.
Terminal state A state that cannot be retried out of: an expired artifact, a revoked link, a revoked agent.
Unclaimed An artifact that belongs to nobody: live, read-only, and expiring 24 hours after creation.