Build the application, not the same infrastructure again.
Every agent application needs a way to find information, remember what matters, use tools, communicate, and decide which actions need human approval. Those capabilities have to work together.
Visca is the developer platform we are building to make that foundation programmable through a coherent API experience. Choose the capabilities your agent needs and build the application around them.
The technology can change. The work should continue.
Models will improve. Tools and devices will change. Developers should be able to adopt those changes without rebuilding identity, memory, permissions, and the rest of the application foundation.
This is the continuity we are working toward: useful context belongs to the person and the application, rather than being trapped in one interface or provider.
The whole agent needs a foundation.
Identity connects the work to an accountable agent. Memory and knowledge supply relevant context. Tools and execution turn a plan into work. Permissions define what is allowed, and communication keeps people involved.
These are parts of the same application, not isolated features. Our platform scope includes all fourteen foundations below, with capability and release boundaries made explicit.
THE COMPLETE ENVIRONMENT
Everything an agent needs
to participate in the real world.
Identity
A durable account, role, history, and place in the world.
An accountable human
A person who grants authority and remains responsible for the decisions that should not be delegated.
Organizational presence
Teams, responsibilities, ownership, and relationships with agents and people.
Communication
An address, inbox, voice, and trusted ability to speak beyond the system.
Time
Meetings, availability, deadlines, recurring responsibilities, and follow-through.
A workspace
One persistent place for context, collaborators, resources, and outcomes.
Compute
A secure computer, browser, sandbox, or isolated environment in which to work.
Files and knowledge
Documents, spreadsheets, diagrams, evidence, media, and shared understanding.
Long-term memory
Continuity across work, people, meetings, decisions, and years.
Connections and permission
Controlled access to systems, applications, data, instruments, and devices.
Financial authority
Budgets, purchasing, subscriptions, invoices, payments, and payouts within human-defined limits.
Models, skills, tools, and training
Capabilities and knowledge that can improve without replacing the agent.
Collaboration with agents and humans
Shared work, review, approvals, handoffs, and history.
Physical embodiment
Sensors, hands, mobile platforms, machines, robots, and eventually a complete physical form.
Your choices should continue, too.
Continuity should preserve a person’s choices, including what the agent must forget. People should be able to inspect context, correct it, revoke access, and decide what follows them into another experience.
A signal is not permission to act. The interface must make an interpretation understandable and leave room to confirm, interrupt, redirect, or stop. New inputs do not silently expand an agent’s authority.
DIGITAL
Code. Research. Operations.
PHYSICAL SYSTEMS
Labs. Instruments. Machines.
EMBODIMENT
Hands. Mobility. Bodies.
Agents can take different forms.
A face on your screen when you want a conversation. A spatial interface when your surroundings matter. A supported silent input when speaking is inconvenient. Different forms should connect to the same ongoing task.
This is what we mean by shape-shifting agents. It is one expression of the broader platform: an interface can change while the application keeps its identity, relevant context, tools, and permissions.
Changing the interface should not mean explaining yourself again. Changing the model should not mean abandoning your history. Changing the vendor should not mean losing the context you built. Universal migration across every device and provider remains an ambition.
Mellow brings the platform into your hands.
Mellow is our personal and development workspace and SDK. People can use the workspace to work with agents; developers can build their own applications on Visca.
Our existing work spans application infrastructure, the native workspace, real-time avatars, signal processing, and EEG acquisition. The first clinical AI customer uses avatar and deployment capabilities.
Existing APIs and SDKs have distinct integration boundaries. The native EEG acquisition SDK is a development preview; broader Signal tooling is planned for Mellow SDK 1.2.x. General neural-intent and subvocal decoding, universal hardware support, and medical claims are not established by acquisition work.
Choose where your context lives.
Private and air-gapped deployments belong within this vision. Mellow is the route to local and private use, with no additional software fee for configuring fully offline workflows. People and organizations should choose where their context lives and where their agents run.
An air-gapped workflow requires local models, assets, tools, and supporting services. Connected providers and remote integrations require connectivity. The deployment must match the experience people actually need.
Build freely. Take the experience further.
We want developers to start freely, experiment, and create experiences we have not imagined. As those applications grow, the platform should grow with them.
Mellow’s workspace is free forever for personal use. Local, private, and fully offline configuration carries no additional Mellow software fee. Hardware, third-party APIs, and optional hosted services or managed support have separate costs. Mellow Glasses and Cap devkits are planned for Q2 2027.
WE ARE BUILDING
One API for everything
your agent needs.
The form can change. The work should continue. The person keeps the choice.
Agents should change shape. Your context should stay yours.