How Agents Access Data in Solace Agent Mesh

Solace Agent Mesh provides multiple approaches to building intelligent agents, but agents are only as valuable as the data they can access. Without connections to your databases, APIs, knowledge bases, and enterprise systems, even the most sophisticated agent remains limited to general knowledge.

This is where data connectors become essential. Connectors bridge the gap between your agents and the information they need to deliver real value. They handle authentication, protocol translation, and data retrieval so your agents can focus on what they do best: understanding user intent and providing intelligent responses.

Understanding Data Connectors

Data connectors in Solace Agent Mesh Enterprise enable agents to access external data sources and services through natural language conversations. You configure each connector once with credentials and connection details for a specific system. Multiple agents can then share the same connector, eliminating the need to duplicate configuration across your deployment.

The connector architecture follows a shared credential model. When you create a connector, you configure it with specific credentials like database passwords, API keys, or service account tokens. All agents assigned to that connector use the same credentials and have identical access permissions to the external system. This design simplifies management but requires careful consideration of access control at the external system level rather than at the connector assignment level.

Types of Data Connectors

Solace Agent Mesh provides four primary connector types, each designed for different data access patterns and integration scenarios.

MCP Connectors

Model Context Protocol (MCP) connectors enable agents to communicate with remote MCP-compliant servers. These connectors provide access to external tools and data sources that implement the MCP standard, which is rapidly becoming a universal protocol for AI agent integrations.

MCP connectors automatically discover the tools that each server provides. When you create an MCP connector, you specify the server URL and authentication method. The connector then fetches the available tools and makes them accessible to your agents. The connector handles all protocol communication, authentication, and request formatting behind the scenes.

The connector supports two transport protocols: Server-Sent Events (SSE) and Streamable HTTP. Authentication options include no authentication for public servers, API key authentication, HTTP Basic or Bearer token authentication, and OAuth2/OIDC flows with automatic token refresh. You can also configure tool selection to control which MCP tools are available to agents, helping you limit scope and reduce complexity.

When to use MCP connectors:

Choose MCP connectors when integrating with third-party services that offer MCP endpoints, connecting to the growing ecosystem of MCP-compatible tools, or when you need a standardized protocol for agent-to-service communication. This is ideal for services like GitHub, Atlassian, Canva, and other platforms that provide MCP servers.

OpenAPI Connectors

OpenAPI connectors allow agents to interact with REST APIs that provide OpenAPI specifications. The connector reads the OpenAPI spec to automatically generate callable tools from API endpoints, handling all the technical details of API integration.

When you create an OpenAPI connector, you upload the OpenAPI specification file and configure authentication. The connector parses the specification to understand the API structure, parameter requirements, authentication methods, request and response formats, and data schemas. It then converts each API operation into a tool that agents can invoke through conversation.

The connector supports OpenAPI 3.0+ specifications in JSON or YAML format. It handles various authentication schemes including API keys sent in headers or query parameters, HTTP Basic authentication with username and password, Bearer token authentication, and OAuth2/OIDC with client credentials flow. The specification file is stored in a public storage bucket where agents can retrieve it during startup.

When to use OpenAPI connectors:

OpenAPI connectors excel when integrating with RESTful APIs that provide OpenAPI specifications, enabling agents to call external services like payment processors or CRM systems, or when you need automatic tool generation from API documentation. This approach works well for services that do not offer MCP servers or any internal API documented with OpenAPI.

SQL Connectors

SQL connectors enable agents to query relational databases using natural language. Instead of requiring users to write SQL queries, agents convert natural language questions into SQL, execute them against your database, and return results in conversational format.

The connector establishes persistent connection pools to your database servers. This pooling architecture improves performance by reusing connections and automatically handles connection lifecycle management, including reconnection after network issues or database restarts.

SQL connectors support MySQL, PostgreSQL, MariaDB, Microsoft SQL Server, and Oracle databases. Each database type uses the same configuration interface but the connector automatically handles database-specific details like SQL dialect, connection protocols, and driver configurations. You simply specify the host, port, database name, and credentials, and the connector takes care of the rest.

When to use SQL connectors:

SQL connectors are perfect when you need agents to answer questions about data stored in relational databases, want to enable non-technical users to query databases without writing SQL, or need to integrate analytics and reporting capabilities into agent conversations. Common use cases include customer data queries, inventory lookups, sales analytics, and operational reporting.

Knowledge Base Connectors

Knowledge Base connectors enable retrieval-augmented generation (RAG) by connecting agents to organizational knowledge repositories. When users ask questions, agents search the knowledge base for relevant information and use that context to provide accurate, company-specific responses grounded in your enterprise documentation.

Currently, Solace Agent Mesh supports Amazon Bedrock Knowledge Bases, which can contain both unstructured documents from sources like S3, web crawlers, Confluence, SharePoint, or Salesforce, and structured data from Amazon Redshift. The connector retrieves information using a consistent API regardless of the underlying data source type.

When you configure a Knowledge Base connector, you provide the knowledge base ID, AWS region, and authentication credentials. The connector can use either static AWS access keys or IAM role-based authentication for AWS-native deployments. You also provide a description that helps the agent understand when to invoke the knowledge base tool.

When to use Knowledge Base connectors:

Knowledge Base connectors shine when grounding agent responses in enterprise documentation and policies, reducing hallucinations by providing factual context from your organization, or enabling agents to answer questions about company procedures, products, or services. This is essential for support agents, HR assistants, or any agent that needs to reference company-specific information.

Creating and Managing Connectors

You create all connector types through the Connectors section in the Solace Agent Mesh web interface. The creation process is straightforward: navigate to the Connectors page, click Create Connector, choose the connector type, and fill in the required configuration fields.

Each connector requires a unique name and connection credentials appropriate for the target system. The web interface guides you through the specific fields needed for each connector type. For example, MCP connectors need a server URL and transport protocol, OpenAPI connectors require an OpenAPI spec file upload, SQL connectors need database host and credentials, and Knowledge Base connectors require a knowledge base ID and AWS credentials.

Once you create a connector, it becomes available for assignment to any agent in your deployment. This reusability is a key advantage. You can connect multiple agents to the same external system without duplicating configuration. When you create or edit an agent through Agent Builder, you simply select from the available connectors to grant the agent access to those data sources.

Access Control

The management of Connectors operations (creating, reading, updating, and deleting) require role-based access  control (RBAC) capabilities. The system provides four granular permissions:

• sam:connectors:create allows users to create new connectors
• sam:connectors:read allows viewing connector configurations and listing available connectors
• sam:connectors:update allows modifying connector configurations and credentials
• sam:connectors:delete allows removing connectors from the system

You can only delete a connector if no agents are currently assigned to it. This restriction prevents breaking deployed agents. If agents use a connector, you must first undeploy those agents and remove the connector assignment before deleting it.

Getting Started

Ready to connect your agents to data? The comprehensive connectors documentation provides detailed setup instructions for each connector type, including prerequisite requirements, step-by-step configuration guides, authentication options, and troubleshooting tips.

Each connector type has its own dedicated guide with specific configuration examples, security considerations, and best practices. Whether you’re connecting to MCP servers, REST APIs, SQL databases, or knowledge bases, the documentation walks you through the entire process. The connector ecosystem is constantly evolving, with new connector types being added regularly to support additional data sources and integration patterns, so stay tuned for updates.

Resources

• Solace Agent Mesh Repository
• Connectors Documentation
• Solace Agent Mesh Documentation
• Have a question? Ask it in the Solace Community Forum