Agents

Cencori Agent Foundry lets you connect any autonomous agent to your infrastructure. All agents include Shadow Mode protection and can be configured, monitored, and controlled from a single dashboard.

Cencori is not the agent — it's the brain. Your agent runs locally (or in the cloud), and Cencori provides model routing, billing, safety controls, and monitoring via a standard OpenAI-compatible API.

Available Blueprints

OpenClaw

The autonomous desktop operator. OpenClaw controls your mouse, keyboard, and browser to automate complex local tasks.

Read full setup guide →

Best for: Desktop automation, file management, browser-based workflows.

# Install
curl -sSL https://openclaw.ai/install.sh | bash
 
# Connect to Cencori
export OPENAI_BASE_URL=https://cencori.com/api/v1
export OPENAI_API_KEY=cake_your_key_here
# Optional: for multi-agent routing overrides
export CENCORI_AGENT_ID=your_agent_id
 
# Run
openclaw onboard

Capabilities:

• Full desktop control (mouse, keyboard, screenshots)
• File system access and management
• Application interaction and automation

[!NOTE] CENCORI_AGENT_ID is optional. Cencori derives the agent from the API key by default.

n8n Workflow

Build custom AI agents visually using n8n's low-code workflow automation platform. Connect to 300+ apps without writing code.

Read full setup guide →

Best for: Multi-app automation, data pipelines, event-driven workflows.

# Install
npx n8n
 
# Connect to Cencori
export OPENAI_BASE_URL=https://cencori.com/api/v1
export OPENAI_API_KEY=cake_your_key_here

Capabilities:

• Visual workflow builder with drag-and-drop nodes
• LangChain integration for advanced AI chains
• 300+ pre-built app integrations (Slack, GitHub, Notion, etc.)

AutoGPT

The original autonomous agent. AutoGPT recursively improves its own prompts and reasoning to achieve complex, multi-step goals.

Read full setup guide →

Best for: Open-ended research, long-running autonomous tasks.

# Install
pip install autogpt
 
# Connect to Cencori
export OPENAI_BASE_URL=https://cencori.com/api/v1
export OPENAI_API_KEY=cake_your_key_here
 
# Run
autogpt run

Capabilities:

• Recursive self-improvement and task decomposition
• Long-term memory across sessions
• Internet access for real-time research

CrewAI

Orchestrate a team of role-based agents that collaborate on complex projects. Each agent has a specialized role and delegates tasks to others.

Read full setup guide →

Best for: Multi-agent collaboration, complex project execution.

# Install
pip install crewai
 
# Connect to Cencori
export OPENAI_BASE_URL=https://cencori.com/api/v1
export OPENAI_API_KEY=cake_your_key_here

Capabilities:

• Role-based agent assignment (Researcher, Writer, Reviewer, etc.)
• Automatic task delegation and sequencing
• Collaborative output synthesis

Python Sandbox

A secure, sandboxed environment for executing Python code generated by LLMs. Ideal for data analysis and visualization tasks.

Read full setup guide →

Best for: Data analysis, code execution, generating charts and reports.

# Install
pip install openai
 
# Connect to Cencori
export OPENAI_BASE_URL=https://cencori.com/api/v1
export OPENAI_API_KEY=cake_your_key_here

Capabilities:

• Isolated code execution with resource limits
• Data analysis with pandas, numpy, matplotlib
• Automatic visualization generation

Custom Agent

Build your own agent from scratch. Any tool that supports a custom OpenAI base URL works with Cencori out of the box.

Read full setup guide →

Best for: Specialized use cases, proprietary workflows, experimentation.

# Connect any OpenAI-compatible agent to Cencori
export OPENAI_BASE_URL=https://cencori.com/api/v1
export OPENAI_API_KEY=cake_your_key_here
# Optional: force a specific agent when needed
export CENCORI_AGENT_ID=your_agent_id

Compatible with: OpenHands, Aider, Continue, LangChain, LlamaIndex, SillyTavern, and any tool that supports custom OpenAI endpoints.

Shadow Mode

All agents include Shadow Mode, a proactive safety layer that:

1. Intercepts tool calls from the LLM before they execute.
2. Inserts them as pending actions in the dashboard Live Feed.
3. Waits for your approval before the agent proceeds.

Toggle Shadow Mode on or off from the agent's configuration page.

Configuration

From the Configuration tab you can:

• Switch Models instantly (GPT-4o, Gemini 2.0 Flash, Claude 3.5, etc.)
• Edit the System Prompt to refine the agent's behavior
• Generate & Roll API Keys for secure authentication
• Delete agents you no longer need