AI Agents

Cainty's AI agent system lets you automate content creation using 6 LLM providers and 24+ models.

How It Works

1. Create an agent — Choose a provider, model, and system prompt
2. Execute — Run manually or on a schedule
3. Review — Generated content enters the queue for review
4. Publish — Approve, edit, or reject from the content queue

Supported Providers

Provider	Example Models	Config Key
Anthropic	Claude Sonnet 4.5, Claude Haiku	ANTHROPIC_API_KEY
OpenAI	GPT-4o, GPT-4o-mini	OPENAI_API_KEY
Google	Gemini 2.0 Flash, Gemini Pro	GOOGLE_API_KEY
DeepSeek	DeepSeek Chat, DeepSeek Reasoner	DEEPSEEK_API_KEY
xAI	Grok 2, Grok 3	XAI_API_KEY
Ollama	Any local model	OLLAMA_BASE_URL

Setting Up API Keys

You can configure API keys two ways:

• .env file — Add keys directly to your configuration
• Admin UI — Go to Admin > Settings > LLM Keys to manage keys through the interface (stored encrypted)

Creating an Agent

Navigate to Admin > Agents > New Agent and configure:

• Name — A descriptive name for the agent
• Provider — Which LLM service to use
• Model — Specific model from the provider
• System Prompt — Instructions that define the agent's behavior
• User Prompt Template — Template for generating content (supports variables)
• Target Category — Where generated posts will be categorized

Content Queue

All AI-generated content goes through a review queue before publishing:

• Pending — Newly generated, awaiting review
• Approved — Published as a post
• Rejected — Discarded

Review content at Admin > Content Queue. You can edit the title, body, and category before approving.

Agent Memory

Agents maintain memory across runs. This helps them:

• Avoid generating duplicate topics
• Build on previous content
• Maintain consistent voice and style

Run History

Every agent execution is logged. View run history at Admin > Agents > Runs to see:

• Execution timestamp
• Provider and model used
• Success or failure status
• Token usage and response time