Business systems
Connections to ERP and CRM, so agents work on real business data.
An agent is useful when it reaches the systems where data and processes live. Connectors link agents to ERP, CRM, databases, APIs and applications already installed at customer sites, even older ones.
Connections to ERP and CRM, so agents work on real business data.
SQL databases, REST APIs and MCP services as standard access points.
Connection to the organization's document systems.
For software houses: bring agentic features into software already in production at customers.
This page describes the available ways to connect. For a specific system, ask us about feasibility and requirements before assuming a ready-made connector exists.
Third-party MCP servers (for example GitHub, Linear, Notion) enter the project as native tools. For Microsoft 365 you start from a template with mail, calendar, OneDrive/SharePoint, Teams and people.
Via APIIf a system exposes an API, a SQL database or a service, a TypeScript function turns it into an agent tool. Credentials stay in the node's secrets, not in the code.
At the customerERP on the local network, internal databases, shared folders: a node runs on a customer machine, next to the systems, without exposing them to the outside.
You write a TypeScript function: every function exported by a connector becomes a tool for the agent, with its real arguments. The code lives in a git repository with commit history.
Yes. A "Connector at your site" node runs on a customer machine (Linux server or Windows PC), next to the local systems. If the machine is off, the functions answer that the site is offline until it reconnects.
Not in the code. They are declared in the node's secrets (a value, an organization secret, a member's personal secret or an OAuth connection) and read at runtime.
It is better not to. Each agent should receive only the privileges it needs, and a public agent should not have access to a customer machine.
Yes, it is one of the intended uses: bringing agentic features into software already installed at customers. The path is on the dedicated software house page.
TypeScript connectors, sandboxes, remote and hosted MCP servers: how to connect agents to business systems.
Making agents work with systems that have no public address: internal ERPs, local databases, shared folders.
MySQL databases, file storage and skill library: the persistent state of agents and code.
Deployment and day-to-day operation of AI Agents built with external tools or with the TeraBrain NX builder.
User, role, permission and security-profile management, with access control.
Activity tracking, secure log retention, monitoring and auditing of the operations performed by agents.
Centralized management of agents and tools that support regulatory compliance, including the applicable requirements of the AI Act.
Legacy-readyRegistration, configuration, control and centralized governance of devices, endpoints, connectors and software resources used by agents.
Integration with company phone systems through SIP Trunk, to enable voice agents connected to the existing telephony infrastructure.
OptionalAn integrated environment to create and configure AI Agents directly inside TeraBrain NX.
Turn your AI Agents into operational, integrated and governed business solutions.