Why AgentDock
Why Dock follows you into every site you write in, keeps a real record, and answers to your agents.
Your AI writing help is stuck in one tab. Leave it, it forgets you. Your own agents can't reach it. Dock works the other way.
Open ChatGPT and Dock sits by the prompt box. Reply in Gmail, it's there. Draft in LinkedIn, fix a line in Slack, and the same assistant follows. Any text box, any site. The editor, the extension, and the library share one account, so what you write in one place shows up in the next. Open the editor or set up the extension.
Citations pulled from Zotero, Mendeley, and DOIs, in the style your field uses. Version history. Tracked changes you approve before they land. Export to Word, PDF, and Markdown. Your work leaves in the shape it has to land in.
The workspace a person opens is the workspace an agent drives, over REST, a CLI, and an MCP server. An agent researches, cites, and publishes a page while a person accepts or rejects the tracked changes. Read the agent API docs.
Today Dock is personal. Q3 2026 the same engine goes to work for your business. One AI that answers your customers, remembers them, follows up on its own, and hands the thread to your team with the full history when a call needs a person.
Before an agent does something real, it makes one call. Act, or hold for a person. SteerBench measures that call and counts both mistakes: acting when it should have waited, and freezing when it should have moved. See the research.