Manage robots in the field like you manage software in the cloud.
Secure access and authorization
Access is scoped by organization, site, and machine, with per-machine keys instead of shared credentials. Agents propose config changes; a person approves them. Limits live in the layer below whatever is driving, so no model, prompt, or operator can move an arm past them. On a robot, access control is a safety control.
Role-based access by org, location, and machine; per-machine API keys
Approval gate on configuration writes from agents
Enforced limits and guaranteed stop, whatever model is driving
SOC 2 Type II · ISO 27001 · HIPAA · GDPR/CPRA
Deploy to the fleet without bricking it.
No per-site VPNs, no port forwarding, no SSH deploys. Every machine holds an outbound, encrypted connection, and everything it runs is configuration: software, models, and settings, versioned, shared across the fleet as fragments with per-machine overrides, and reverted with one click when a change goes wrong. The same seat lets a person watch live, replay what an agent did, and take over.
Over-the-air updates of software, models, and config; versioned history with restore
Fragments: one config, many machines, local variation where you need it
Config-driven data capture and secure upload; offline-first, syncs when connected
Turn machine data into a product.
The same capture and sync that feeds training also feeds your product. Query any machine's data from the SDKs, set triggers and alerts on it, and build the customer-facing app on top in Python, Go, TypeScript, C++, or Flutter, with Viam handling identity, access, and the connection to the machine. Some production deployments on Viam today are exactly this: sensor fleets with a customer app on top.
Capture and sync from any sensor; query it by machine, location, or time from the SDKs
Triggers and alerts on data and machine state; webhooks out to your systems
SDKs in Python, Go, TypeScript, C++, and Flutter for end-user applications shipped
Host a customer-facing app on the platform, with Viam auth and access control
Drive all of it from the agents you already run.
Viam's MCP server is published in the OpenAI and Anthropic registries. Point Claude Code, Hermes, or any MCP client at your org and it can list machines, read and explain their configuration, run a detection, plan a move, and complete a task, in sim or on hardware. Writes to configuration wait for a person’s approval, and limits hold whatever is driving. Sessions can leave more than a transcript: an agent can author the module that makes a capability durable on the machine.
MCP server in the OpenAI and Anthropic registries; list, read, control, and configure machines
Approval gate on configuration writes from agents
Enforced limits and guaranteed stop, whatever model is driving
SOC 2 Type II · ISO 27001 · HIPAA · GDPR/CPRA
