Agents
Agents are autonomous systems that understand natural language and use tools and reasoning to accomplish tasks.
- Create an agent: Define the agent's name, description, and available tools.
- Write instructions: Create templates that guide the agent's behavior.
- Configure tools: Set up corpus access permissions and any external integrations.
- Test agents with sessions: Start conversations and iterate on your configuration.
To build your first agent, see the Agent quick start. Or, check out our Agents APIs.
Concepts
The core concept to understand about agents is that their behavior is defined by instructions. The agent uses these instructions alongside information from a conversation session to determine how to respond to user input, including which tools to use.
These are the other core concepts when it comes to agents:
- Tools: Tools provide agents with capabilities to interact with data and external systems.
- Sessions: Sessions preserve context throughout a conversation so the agent can consider prior information when responding to a query.
Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an emerging and rapidly changing standard for communication between agents and tools. Introduced by Anthropic and now adopted across multiple agentic ecosystems, MCP defines a structured way for agents to invoke external tools, pass contextual input, and receive structured responses. MCP acts as a universal connector between AI agents and enterprise services.
In the tech preview of Agents, Vectara only supports MCP as a client. This means Vectara agents can call MCP-enabled tools and services hosted on external MCP servers.
You can run your own MCP server to expose your tools and data within the Vectara platform. To help you get started, we provide an unsupported open-source MCP server implementation. Please note that while you can connect your self-hosted MCP server to Vectara, you are responsible for its management and maintenance.
Why MCP is important
MCP eliminates the need for custom integrations by providing a universal protocol that any AI system can use to access external capabilities. Instead of building custom integrations for every tool or service, MCP creates a standardized interface that works across different LLM providers and enterprise systems. MCP provides the following benefits:
| Benefit | Description |
|---|---|
| Standardized tool access | Access tools through a common schema and interface. |
| Dynamic tool discovery | Agents can discover tools available at a registered MCP server. |
| Enterprise integration | Tools can represent internal APIs, RAG systems, databases, or even other agents. |
| Auditability and security | Permissions and access control can be managed per tool, enabling secure orchestration. |
| Model-agnostic | MCP works across agent platforms (Vectara, Claude, OpenAI SDK, Google Gemini, and so on). |