Managing Your AI Agents
AI agents are the core of your RightClaw instance. Each agent is a configurable AI assistant that connects to a model, has its own workspace, can be assigned skills, and responds across your connected messaging channels. The Agents section of the dashboard is where you manage them.
Viewing Your Agents
Navigate to Agents from the sidebar to see all configured agents on your instance. Each agent appears as a card showing key information:
- Status — Online, offline, or busy. Online means the agent is running and ready to accept messages. Busy means it is currently processing a request. Offline means it is stopped or misconfigured.
- Model — The AI model the agent is configured to use (e.g., Claude Sonnet, GPT-4o, DeepSeek Coder).
- Workspace — The file directory the agent operates within.
- Last activity — When the agent last processed a message or performed an action.
- Session count — How many conversations the agent has handled.
- Latency — Response time metrics for the agent.
Agent Details
Click on any agent card to open its detail view. Here you can see:
- Configuration — The agent's system prompt, model selection, API key association, temperature, and other parameters.
- Skills — Custom skills assigned to the agent. Skills extend what an agent can do — for example, web searching, file operations, or API integrations.
- Stats — Historical performance data including total sessions, messages processed, tokens consumed, and average response time.
- Connected channels — Which messaging platforms this agent is wired to respond on.
Agent Configuration
Each agent's configuration determines its behaviour:
- System prompt — Instructions that define the agent's persona, role, and constraints.
- Model — Which AI model to use. Different models have different strengths, costs, and capabilities.
- Temperature — Controls response creativity. Lower values produce more focused, deterministic outputs. Higher values produce more varied, creative responses.
- Skills — Optional capabilities that extend the agent's toolset beyond basic conversation.
- Workspace — The directory the agent can read from and write to for file-based tasks.
Creating New Agents
Creating a new AI agent requires admin access on your RightClaw instance. If you have admin permissions, you can add agents from the Agents section by clicking the create button and configuring the agent's name, model, system prompt, and connected channels.
If you do not see the option to create agents, your plan may not include admin access or it may need to be enabled. Contact support@rightservers.com for assistance.
Monitoring Agent Health
The Agents section also helps you spot issues quickly. If an agent shows as offline, check:
- Whether the connected API key is valid and has sufficient credits.
- Whether the agent's workspace path exists and is accessible.
- Whether any connected messaging channels have lost their authentication.
The dashboard's Activity section provides a detailed event log that can help diagnose agent issues, including error messages and timestamps.
