Expand AI capabilities with
Model Context Protocol

Connect external tools, RAG collections, and complex workflows to your AI agents via the unified MCP protocol. Secure interaction between neural networks and your data without writing hundreds of lines of integration code.

Create first agent → API Documentation

Open standard for LLM integration

AI Agent (LLM)
Theta AgentEngine
MCP Protocol
RAG Collections
Vector Databases
External APIs
Integrations
Internal Systems
Custom MCP

What is Model Context Protocol (MCP) and why is it needed?

Developers of AI applications need a secure, standardized way to interact with the outside world and corporate data. Model Context Protocol (MCP) is an open standard that solves this problem, turning the chaos of API integrations into a single universal protocol.

Isolated Configuration (Plug & Play)

Manage MCP server settings, endpoints, and access through a single standardized JSON format.

Feature

Unified management format

Hot Attach / Detach

Bind and disconnect tools to language models on the fly without needing to rebuild or restart the agent.

Flexibility

Connect on the fly

Granular Access and Security

Strict permission control at the Workspace level and Tenant isolation to prevent data leaks.

Security

Enterprise-grade protection

Use Cases

Thanks to MCP protocol support, your agents can perform real-world tasks and automate processes:

Code and Tracker Integration

Read GitHub/GitLab repositories, analyze pull requests, and manage Jira tasks directly through the agent.

View integration

Business Data Access

Execute SQL queries to internal databases or get context from CRM systems (Salesforce, HubSpot) in real-time.

View integration

Workflow Automation

Send notifications to Slack, manage files, call webhooks, or run local scripts.

View integration

Platform Core Concepts

RAG Collections (Resources & Data)

Connect vector databases to provide agents with up-to-date context (Retrieval-Augmented Generation). Turn your documents into AI-accessible resources.

Action

Configure RAG integration →

Tools (Instruments & Actions)

Integration with external APIs, running local scripts, sending webhooks, and direct interaction with microservices. The agent decides when to call the right tool.

Action

MCP Tool Catalog →

Workflows (Scenarios)

Call chains and multi-step processes where the AI agent participates step-by-step, making decisions based on data received from MCP.

Action

Create AI workflow →

Bring Your Own MCP (BYO MCP)

You are not limited to built-in tools. Our platform allows you to connect any custom MCP servers compatible with the open standard.

1

Deploy MCP Server

Run your server locally or in the cloud using official MCP SDKs for Python, TypeScript, or C#.

2

Create New Resource

In the Workspace control panel, add a new resource of type Tool or Other.

3

Specify Settings

Set the URL (Endpoint) of your server, custom headers, and authentication parameters in JSON format.

Custom MCP Creation Guide
custom-mcp-config.json
{
  "mcpKind": "Tool",
  "endpoint": "https://your-custom-mcp-server.com/v1/execute",
  "auth": "Bearer <YOUR_SECRET_TOKEN>",
  "customHeaders": {
    "X-Tenant-Id": "tenant-123"
  }
}

Management via API (For Developers)

REST API Quick Start
# Get agent MCP list
GET /api/v1/agents/{id}/mcps?workspaceId={...}

# Attach MCP to agent
POST /api/v1/agents/{id}/mcps?workspaceId={...}

# Detach tool
DELETE /api/v1/agents/{id}/mcps/{mcpId}

Lifecycle and Data Model

Platform Entities

Mcp (basic tool configuration) and AgentMcp (connection to a specific AI agent).

Creation and Configuration

Create an MCP resource within a Workspace. Define ConfigJson (connection parameters, credentials, routing).

Attachment (Attach)

Link the resource to an Agent via a many-to-many relationship. Manage tool attachments via our REST API for CI/CD automation.

Security and Access Control

Strict Isolation

All tools are isolated within the Tenant and Workspace. Unauthorized access between sandboxes is completely excluded.

Status

Secure

One-Click Management

Instantly enable and disable integrations using the IsEnabled: true/false flag. You have full control over agent access.

Status

Active

Secret Protection

Secure storage of API keys, tokens, and credentials in encrypted configurations (they are not passed in plain text to language models).

Status

Encrypted

Ready to expand your AI agents' capabilities?

Start using Model Context Protocol now and connect your first RAG database or API in 5 minutes.

Create agent for free