Control AI access
Reduce shadow IT by giving your business one governed place to approve apps, providers, keys, and model access.
Connect AI apps to local and online models, choose which MCP tools they can use, and review usage in one place.
Use Nexus to connect your apps to the models you choose. Set model access, manage app tokens, review requests and usage, and add routing and protection controls when you need them.
Reduce shadow IT by giving your business one governed place to approve apps, providers, keys, and model access.
Keep sensitive data private and secure on local infrastructure, while still using online models when the job needs them.
Apply extra protections like scoped tokens, PII redaction, key blocking, and routing rules before requests reach models.
Nexus gives teams and power users one place to connect AI models and MCP tools, protect sensitive data with guardrails, and keep provider credentials out of every application.
Bring local runtimes, hosted providers, and MCP servers into one Nexus-managed gateway.
Standardize model settings and apply Data Protection, PII redaction, key blocking, and routing rules before requests reach models.
Give each client token only the models and MCP tools it needs, with allow, ask, or block policies for tool calls.
Nexus routes inference, applies configured protections, and records model and tool activity without exposing provider credentials.
Nexus is more than a gateway. It can choose the right model path, distribute traffic, and coordinate inference capacity so every request does not land on the most expensive model.
Send each request to the right model for the job. Keep expensive models for work that needs them, and route routine tasks to cheaper or local models to reduce token costs.
Distribute requests across available runtimes and machines so teams can keep throughput steady without pointing every app at one overloaded endpoint.
Coordinate multiple machines as shared inference capacity for larger local workloads, central servers, or teams that need more horsepower behind the gateway.
Use Nexus as the operating layer for local runtimes, hosted providers, governed app access, and practical usage visibility.
Give applications one governed catalog across local runtimes and hosted providers, including AWS Bedrock and Azure AI.
Connect tool integrations through Nexus so approved clients can discover and call tools with scoped permissions, approvals, and activity logging.
Keep provider credentials in one controlled place instead of copying them across tools.
Connect approved applications through familiar OpenAI- and Anthropic-compatible endpoints.
Manage supported Ollama, llama.cpp, and MLX runtimes alongside hosted model access.
Issue scoped tokens per application instead of sharing one master key across every integration.
See requests, latency, and usage by application and provider from one place.
Chat with models, presets, routes, and load-balanced paths from inside Nexus, then benchmark performance before routing production traffic.
Review and approve new model or provider connections before applications can use them.
Keep the gateway off the LAN by default and open access only when you choose to.
Run Nexus across your team’s macOS, Windows, and Linux devices, operate your shared models on private infrastructure, and reach approved cloud providers through a protected route only when your work calls for them.
Explore Enterprise deploymentsNexus is available as a free download. Pro and Enterprise add managed scale, protection controls, logging, and team governance.
Download Nexus and run the local gateway on your machine.
Managed Nexus control for advanced access, protection, and observability.
You can purchase Pro from the Nexus Console.
Download NexusShared controls for small teams through larger enterprises, stricter environments, and managed deployments.
| Feature / Attribute | Free | Pro | Enterprise |
|---|---|---|---|
| Usage & Billing | |||
| Gateway Requests | Unlimited | Unlimited | Unlimited |
| Managed machines | 1 | 3 | Starting at 25 included |
| Client tokens | 10 | 20 | Starting at 25 included |
| Deployment | |||
| Operating systems | macOS, Windows & Linux | macOS, Windows & Linux | macOS, Windows & Linux |
| AWS Bedrock & Azure AI providers | Not included | Included | Included |
| Administration | |||
| Desktop Console | Included | Included | Included |
| Cloud console | Not included | Included | Included |
| Teams / RBAC | Not included | Not included | Included |
| Team / User rate limit | Not included | Not included | Included |
| Single Sign-On (SSO) | Not included | Not included | Included |
| MDM deployment & AI governance | Not included | Not included | Included |
| Routing & Safety | |||
| Data Protection | Not included | Included | Included |
| Guardrails | Not included | Local policies | Organization-wide policies |
| Smart Routes | 3 | Included | Included |
| Load Balancer | Not included | Included | Included |
| MCP tools gateway | Allow or block all tools | Tool profiles and rules | Team tool profiles |
| Cluster | Not included | Not included | Included |
| Fallbacks | Not included | Included | Included |
| Logging & Support | |||
| Basic logging | Included | Included | Included |
| Team/User logging | Not included | Not included | Included |
| Priority Support | Not included | Not included | Included |
The number of machines or inference runtimes registered with Nexus to provide model inference, such as employee Macs, Mac Studios, or centralized servers.
Client devices do not automatically count as managed machines. A laptop simply consuming inference from a central server is a client; it only counts as a managed machine if it also provides inference through Nexus.
Humans and AI workloads are treated equally. A client can be a person, AI agent, application, workflow, or service.
Published plan capacity is based on the number of managed machines registered with Nexus.
No. Nexus does not charge based on the number of tokens processed and has no gateway request limits, so heavy AI usage does not create per-request fees or overages.
Nexus is designed for local and centralized AI. A deployment can range from a few shared inference servers to hundreds or thousands of employee devices contributing local inference.
Yes. Capacity can be enforced locally, allowing Nexus to support private and air-gapped environments without requiring usage telemetry to Msty.
Download the desktop gateway for the machine that runs Nexus and routes approved apps, models, credentials, and local runtimes.