Core Concepts
Understanding the fundamental building blocks of CogTog will help you create powerful AI agents and workflows.
Agents
Agents are fully autonomous AI entities that think, plan, and execute complex tasks independently. They break down requests, make decisions, and deliver results without hand-holding.
- Agents can use any LLM (OpenAI, Anthropic, Google, Ollama)
- Memory nodes maintain context across conversations
- Specialized agents for different tasks can work together
- Agents can invoke tools and take real actions
Orchestrator
The Orchestrator coordinates multiple agents to work together on complex tasks. It manages agent handoffs, routes tasks to specialized agents, and combines their outputs into cohesive results.
- Coordinate multiple agents for complex workflows
- Route tasks to specialized agents automatically
- Manage agent handoffs and data passing
- Combine outputs from multiple agents
Computer Use
Agents can see your screen, click buttons, type text, and interact with any application. Full desktop automation at your command.
- Screen capture and visual understanding
- Mouse and keyboard control
- Interact with any desktop application
- Automate repetitive GUI tasks
Nodes
Nodes are the building blocks of your workflows. Each node performs a specific function, from receiving input to processing data to generating output.
- Input nodes receive data (text, files, API calls)
- Processing nodes transform and analyze data
- Output nodes deliver results (text, files, webhooks)
- Control flow nodes manage branching and loops
Workflows
Workflows connect nodes together to orchestrate complex agent behaviors. Data flows through connections from node to node, enabling sophisticated multi-step automation.
- Drag and drop nodes onto the canvas
- Connect outputs to inputs to create data flow
- Workflows can be nested inside other workflows
- Export workflows as reusable templates
Chat Interface
Test your agents with a built-in chat interface. Have conversations, share results with stakeholders, and gather feedback in real-time.
- Interactive testing environment
- Share conversations with team members
- Gather feedback from stakeholders
- Debug agent responses in context
Knowledge Base
Connect documents, databases, and APIs to create RAG-powered agents with rich context. Your agents can search and retrieve relevant information.
- Vector database integration for semantic search
- Index documents, PDFs, and web content
- Automatic context retrieval for agent prompts
- Connect to external databases and APIs
Tools
Tools extend agent capabilities by allowing them to perform actions like web searches, file operations, API calls, and database queries.
- Built-in tools for common operations
- Create custom tools for your specific needs
- Connect to external APIs and services
- MCP (Model Context Protocol) server support
Team Collaboration
Work together in real-time with your team. Share agents, templates, and configurations. Collaborate on complex projects seamlessly.
- Share agents and workflows with teammates
- Collaborative editing and review
- Shared template library
- Role-based access control
Version Control
Built-in git integration for managing your agent configurations. Branch, merge, and track changes to workflows. Use worktrees for parallel development.
- Git integration for version history
- Branch workflows for experimentation
- Worktrees for parallel development
- Merge and track configuration changes
Execution
When you run a workflow, CogTog executes nodes in order based on their connections. Watch execution in real-time, see agent reasoning, and debug any issues.
- Real-time execution visualization
- Step-by-step debugging mode
- Variable inspection at each step
- Execution history and logs
Templates
Start faster with pre-built templates for common use cases like chatbots, RAG systems, and task automation. Create and share your own templates.
- Pre-built templates for common use cases
- Chatbots, RAG, and automation templates
- Create custom templates from workflows
- Share templates with your team
How It All Works Together
When you build an AI agent in CogTog, you're creating a workflow that orchestrates how your agent processes information and takes action.
You start by adding nodes to the canvas. These nodes might include an input node to receive user messages, an LLM node to process them with AI, and an output node to return responses.
Connections between nodes define how data flows through your workflow. When you run the agent, CogTog executes each node in order, passing data along the connections.
For complex tasks, the Orchestrator coordinates multiple agents working together. It routes tasks to specialized agents, manages handoffs, and combines their outputs into cohesive results.
Tools and Computer Use extend what your agents can do. Agents can search the web, query databases, call APIs, interact with desktop applications, and automate tasks.
The Chat Interface lets you test agents interactively, while Team Collaboration features let you share and iterate with your team. Use Version Control with git and worktrees to manage changes safely.