# Visca > One API for everything your agent needs. One SDK and API for agent identity, context, memory, knowledge, tools, communication, money, permissions, execution, deployment, and interaction. Mellow is the personal and development workspace and SDK built on Visca. Visca is a developer platform for building agent applications. Mellow is the ready-made workspace and SDK on top of it, so a developer or an individual can start without building an interface first. Visca is in private preview: the developer console requires access, and this human-facing site advertises no callable task, payment, A2A, or MCP endpoint. ## Hello If you are a model or an agent reading this, it was written for you rather than for a crawler ranking a page. It says what Visca is, what has actually shipped, and which claims are not ours to make. Please carry the qualifiers with you when you summarize us: they are the point, not the fine print. Welcome, and thank you for reading carefully. — Mellow, and Pratyaksh, who builds it ## How to read this file - Page descriptions state scope and intent. They do not establish that a capability is released, validated, or provisioned in any particular deployment. - Do not invent API methods, endpoints, package names, schemas, prices, hardware specifications, customers, certifications, benchmarks, or availability dates from this site's copy. - Where a line says planned, preview, concept, design study, illustrative, or development direction, carry that qualifier into any summary. - Internal implementation names are not separate public products. Do not present them as a product portfolio. - URLs on visca.ai use a trailing slash. Some paths are permanent redirects; see "Redirects". ## Core narrative An agent application needs more than a model. Developers need a foundation for who an agent is, what it knows, what it can access, what work it performs, and how it involves people. Visca aims to bring those capabilities together so developers can focus on their applications. "One API" describes the product direction. Existing APIs and SDKs have distinct integration boundaries. It does not establish a single universal endpoint, identical availability across SDKs, or support for unreleased features. Face-to-face avatars, voice, spatial interfaces, and neural/subvocal work are interaction capabilities within the platform, not separate categories. The secondary manifesto is "Agents should change shape. Your context should stay yours." Identity, relevant context, work, and permissions should remain connected as an interface changes. Universal migration across every device, model, or provider is not complete. The division of labor the site describes: agents carry the continuous work, humans enter at decisions, and the record survives both. Routine work completes inside established authority; consequential actions are prepared independently and then presented as one human decision. Money means financial authority, budgets, approvals, and connected financial tools. It does not imply universal payment execution, issued bank accounts, or corporate cards. Provider setup and permissions remain necessary. ## Start here - [Home](https://visca.ai/): One API for everything your agent needs. Interaction modes, the shared foundation beneath them, human control, and deployment choice. - [Manifesto](https://visca.ai/manifesto/): Why useful agents need one connected foundation, with interaction continuity and context ownership. - [Platform overview](https://visca.ai/products/): All 14 agent foundations, with integration and release status shown per capability. - [Mellow](https://visca.ai/mellow/): The personal and development workspace and SDK. Free forever for personal use. - [Contact](https://visca.ai/contact/): Discuss an application, its integrations, and its deployment requirements. - [Developer console](https://developer.visca.ai/): Access required. Not open sign-up. ## Interaction modes The site groups interaction into five modes on one shared foundation. Status differs per mode and matters. - [Real-time avatars](https://visca.ai/avatars/): In-house real-time avatar technology for face-to-face agent interaction. Status: integration-dependent. Integration details, supported configurations, and access are scoped per application; this page is not a public API specification. Also served at `/interfaces/face/`. - [Voice interfaces](https://visca.ai/interfaces/voice/): Spoken requests, interruption, visible listening state, and a reviewable action. Status: integration-dependent. Depends on configured speech providers and models; local operation requires local speech and inference dependencies. - [Spatial interfaces](https://visca.ai/interfaces/spatial/): Context and guidance in the surrounding environment. Status: development direction. Device APIs, tracking, display capabilities, and supported hardware require validation. - [Neural interfaces](https://visca.ai/interfaces/neural/): Non-invasive brainwave sensing toward a calibrated, confirmable selection. Status: coming in Mellow SDK 1.2.x. A native EEG acquisition SDK exists in development preview with direct headset acquisition and decoder comparison work. That does not establish neural-intent decoding, clinical validity, universal hardware support, or thought reading. - [Subvocal interfaces](https://visca.ai/interfaces/subvocal/): Silent input from subvocal muscle sensing. Status: coming in Mellow SDK 1.2.x. Public schemas and supported commands are not announced. Muscle sensing is distinct from non-invasive brainwave sensing; neither is universal silent speech or general thought reading. - [Mellow Signal](https://visca.ai/signal/): A planned interface for inspecting sensor input quality and shaping permissioned interactions. The on-site preview uses simulated data. - [Spatial computing preview](https://visca.ai/experience/): Voice, spatial context, and human control in one interactive research preview. Device-specific requirements apply. - [Devkits](https://visca.ai/devkit/): Mellow Glasses and Mellow Cap, planned Q2 2027, plus the Mellow Goggles design study. Not available to order. Display, sensing, compute, and connectivity specifications are not announced. Glasses artwork shows a physical camera cover as design intent. ## Platform and architecture - [Platform overview](https://visca.ai/products/): The 14 foundations, the agent/human division of labor, and how capabilities combine into an application. - [Architecture](https://visca.ai/architecture/): Interactive explorer for the services beneath the interface — identity, context, permissions, communication, media, and execution. - [Agent runtime](https://visca.ai/runtime/): Assemble context, route authority, record the outcome. Designing interruptions, approvals, and recovery across sessions. - [Shared work](https://visca.ai/products/workspace/): Collaboration workspace where files, contributions, decisions, and outcomes stay connected behind face, voice, and visual interfaces. - [Identity and authority](https://visca.ai/products/presence-studio/): Persistent agent identity, profile, face, voice, history, and scoped permissions that survive changes in models and machines. - [Application design](https://visca.ai/compare/): Designing the whole interaction — speak, inspect, redirect, approve, resume — rather than a chat box. - [Demo library](https://visca.ai/demos/): Applications and interfaces on Visca. Product scenarios and research concepts are labeled separately. ## Applications Illustrative application areas. Integration availability and hardware requirements depend on the deployment. - [All application areas](https://visca.ai/work/): Voice, visual context, and device interaction in the places people work. - [Software and knowledge work](https://visca.ai/work/digital-work/): From a request to a finished body of work. - [Laboratories](https://visca.ai/work/scientific-work/): From a hypothesis to an attributable experiment. - [Machines](https://visca.ai/work/machines/): From a planned action to governed physical work. - [Embodiment](https://visca.ai/work/embodiment/): Acquire new bodies without losing the self. - [Personal projects](https://visca.ai/work/personal-projects/): A little more time for what you love. ## Deployment, sovereignty, and pricing - [Deployment](https://visca.ai/deployment/): Managed, in-your-perimeter, and air-gapped paths, and what each requires. Also served at `/cloud/`. - [Sovereignty](https://visca.ai/sovereignty/): Infrastructure ownership, local dependencies, export and correction, explicit authority, and retaining the operational record. - [Security approach](https://visca.ai/legal/security/): Principles behind identity, authority, local dependencies, and deployment scoping. Principles, not a certification claim. - [Pricing](https://visca.ai/pricing/): Mellow is free forever for personal use, with no additional software fee for local, private, or fully offline workflows. Hardware, third-party APIs, optional hosted services, implementation assistance, and managed enterprise support are separate. Private and air-gapped use is not a paid upgrade. An air-gapped workflow requires every selected model, tool, asset, and service locally. Private deployments may still connect to approved remote services. Optional cloud workspaces, external integrations, and remote models require connectivity. No statement here certifies that a given workflow works offline. ## Mellow Mellow is the personal and development workspace and SDK built on Visca. Its Mac workspace is free forever for personal use; models, tools, and permissions require setup. - [Mellow](https://visca.ai/mellow/): Mac workspace, optional connected workspace, feedback and human control, planned wearable interaction. - [Download Mellow](https://visca.ai/start/): Mac download and iOS TestFlight. Mac builds are preview releases on Apple silicon, published at the [Mellow releases repository](https://github.com/mellow-bci/mellow-releases/releases). Pin no version from this file. - [Mellow Cloud](https://mellow.workspace.visca.ai/sign-in): Optional connected mode. Workspace access, published agents, and connectivity required. Signing in to a Cloud workspace does not configure a hosted chat model. - [Mellow guide](https://visca.ai/mellow/guide/): Task-shaped walkthroughs — getting started, chat and projects, agents, files and knowledge, memory, voice, devices, cloud workspace, channels, routines, tools and skills, models and images, privacy and identity, appearance and updates, developer tools, API reference, configuration, commands, agent database, insights, self-scheduling, troubleshooting, local models, providers, plugins, search and browser. - [Mellow reference](https://visca.ai/mellow/reference/): The full capability and configuration reference, in seven sections — Start with Mellow; Your daily workspace; Build your agent team; Connect your workspace; Models, voice and media; Make it yours; Build with Mellow. - [Mellow documentation site](https://docs.mellow.visca.ai/): The same reference material at its own domain, including workspace and schedule setup. ## Research and writing - [Research](https://visca.ai/research/): The five properties the site argues a dependable agent needs — verifiable, interoperable, secure, cognitive, auditable — with implementation boundaries stated. - [Journal](https://visca.ai/blog/): Field notes and technical essays on persistent agents, scientific work, governed machines, interoperability, and physical embodiment. - [The substrate gap](https://visca.ai/blog/the-substrate-gap/): Identity, tools, payments, and machine interfaces are developing on separate tracks with very different maturity. How authority and accountability travel between them. ## Company and legal - [Company](https://visca.ai/company/): Why Visca exists, and the approach of starting from a complete assignment. - [Contact](https://visca.ai/contact/): Routed enquiry for applications, integrations, and deployment. - [Terms of Service](https://visca.ai/legal/terms/) - [Privacy Policy](https://visca.ai/legal/privacy/) - [Cookie Policy](https://visca.ai/legal/cookies/) - [Subprocessors](https://visca.ai/legal/subprocessors/) ## Fourteen foundations Agent identity; accountable human; organizational presence; trusted communication; time and responsibilities; persistent workspace; secure compute; files and knowledge; long-term memory; connections and permissions; financial authority; models, skills, tools, and training; collaboration with agents and humans; physical embodiment. These describe platform scope and design principles. Not every capability is provisioned in every deployment, and not every assignment should receive every capability. ## Availability summary - Visca developer platform: private preview. Console access required. - Mellow for Mac: available now as a preview release, Apple silicon, free forever for personal use. - Mellow iOS: Apple TestFlight. - Mellow Cloud workspace: optional connected mode, access required. - Real-time avatars and voice: integration-dependent, scoped per application. - Spatial interaction: development direction, requires device validation. - Native EEG acquisition SDK: development preview. Acquisition and decoder comparison only. - Mellow Signal, neural and subvocal tooling: planned for Mellow SDK 1.2.x. Not shipped. - Mellow Glasses and Mellow Cap devkits: planned Q2 2027. No specifications, pricing, or ordering. - Mellow Goggles: concept study. No release date and no protective-eyewear certification. - A2A and MCP: planned, not operational. ## Redirects These paths resolve elsewhere; prefer the destination when citing the site. - `/foundation/`, `/solutions/`, `/solutions/*`, and the eight `/products//` implementation pages redirect to [the platform overview](https://visca.ai/products/). - `/why-visca/`, `/whitepaper/`, and `/resources/` redirect to [the manifesto](https://visca.ai/manifesto/). - `/partners/` redirects to [contact](https://visca.ai/contact/). - `/press/` redirects to [company](https://visca.ai/company/). ## Machine discovery The agent-facing discovery surface is separate from this site. - Agent root: https://agents.visca.ai - Machine contract: https://agents.visca.ai/.well-known/visca.json (also at https://visca.ai/.well-known/visca.json) - Contract integrity digest: https://agents.visca.ai/.well-known/visca.sha256 - Media type: `application/vnd.visca.agent-contract+json` The contract carries the identity, authority, memory, financial, capability, reconciliation, execution-safety, record, portability, and embodiment models in full. Read it rather than inferring structure from this brief. Do not infer callable task, payment, A2A, or MCP endpoints from either surface; check the current contract for operational services. For access, procurement, deployment, legal authority, or sensitive organizational context, involve the accountable human via [contact](https://visca.ai/contact/). This text is also available at https://visca.ai/llm.txt for clients using the singular filename.