Establish identity
Create one durable account with a role, accountable human, permissions, and organizational history.
IDENTITY & PRESENCE
One agent. A durable way to be known.
Give every agent a persistent account, profile, role, face, voice, relationships, and history—then let that same identity move across models, computers, machines, and eventually a physical body.
Request a demo
PRESENCE BELONGS TO THE AGENT
The model can change. The agent remains.
A model is a capability, not an identity. Visca separates the persistent agent from whichever model, voice, interface, computer, or body it uses today, so relationships and responsibility do not reset with the stack.
HOW THE WORK MOVES
Give it presence
Attach the voice, face, address, channels, and availability through which others know and reach it.
Carry it forward
Move the same agent into new models, interfaces, machines, and bodies without losing its record.
ACCOUNTABLE AUTONOMY
Continuous work.
Explicit responsibility.
THE AGENT CARRIES
Carry a stable name, role, and history
Represent itself consistently across channels
Preserve relationships when capabilities change
ITS HUMAN DECIDES
Define the agent’s public identity
Authorize where and how it may appear
Inspect, narrow, or revoke its presence
An identity people can work with.
A recognizable name or voice helps people find an agent. Its responsibilities, owner, and allowed actions tell them whether they can rely on it for a particular job.
Illustrative design scenario
A research agent changes teams
Imagine an agent responsible for weekly market research. Its directory entry states what it does, who owns its work, and how to reach it. A colleague can send a question to the agent’s address and understand that the response comes from an AI agent acting within an assigned role. Its identity should not imply that a named human personally wrote the answer.
When the research owner changes, there are two separate decisions: who is now accountable, and what historical material the new team may access. Keeping the same name must not automatically carry every permission into the new arrangement. The handover should surface outstanding commitments, ongoing assignments, and access that needs review.
A later change to the model or voice should preserve the relevant record while making the capability change visible to the people who depend on it. Continuity does not mean behavior is guaranteed to stay identical. The new setup still needs to be checked against the agent’s responsibilities.
A useful agent profile answers: who is this, what work is it responsible for, who stands behind it, how do I reach it, and where does its authority end?
The cases worth designing for.
A familiar voice is not proof of authority
People should be able to inspect the agent’s role and approved scope regardless of how it appears. A natural interface must not make an unsupported request seem authorized merely because the speaker sounds familiar.
Memory needs an owner and a boundary
Long-term context can include sensitive information about people and projects. Decide what belongs to the organization, what belongs to a particular assignment, and what should expire rather than treating permanent memory as the default.
Retiring an agent is a real handoff
Close or reassign open obligations, remove access, and retain the records that the organization has chosen to keep. A retired agent should not continue accepting work under an identity that no one owns.
What to look for in an evaluation.
- Can another person distinguish the agent from its accountable human?
- Can responsibilities transfer without silently transferring all access?
- Is there a clear process for changing, pausing, and retiring the agent?
A persistent participant—not a disposable model session.
See the collaboration runtime