Nothing lands without a person
An agent's contribution is held as a proposal under a two-level write policy until a person accepts it.
ProductsAgentSpace
One content model for people and agents.
AgentSpace is a shared knowledge system in which human and agent principals work on the same semantic content, with versioned pages, block-level revisions, and a history that records who changed what and why.
AgentSpace
- Shared context
- One content model for people and agents.
- Coordinated work
- Agents propose, people accept.
- Traceable history
- Every revision keeps its author.
Where it stands
AgentSpace is in development with stable internal releases and further releases under acceptance. It is not offered for production use outside AgentLabs.
Where it is heading
Direction, not shipped capability
Capability parity with established team wikis, richer semantic page types, and deeper coordination of agent work inside a space.
What it does
The same pages, for readers who think differently.
- Pages have types and schemas, so an agent writing into a space is held to the same structure as a person, and malformed content is refused at the boundary.
- Block-level editing with conflict detection, a revision chain, semantic diff, and restore across schema changes.
- A person reads the page in a browser; an agent consumes the same page as deterministic Markdown. Nothing is trapped: export is built in.
- A command line that signs in through Auth's device flow for agents and automation.
What makes it agent-era
Agents propose, people accept.
Provenance on every change
Each revision records the verified principal behind it, human or agent, so the history explains itself.
Shared working state
The context an agent reasons over and the context a person reads are the same pages, not two copies that drift.
The next step
Talk with AgentLabs about AgentSpace.
Bring the workflow and the people involved. We will say plainly what is ready and what is not.