Understanding Agents
What agents are in Deslicer, how they differ from generic chatbots, and the agent lifecycle
Understanding Agents
Deslicer AI agents are conversational assistants equipped with tools and connected to your integrations. They differ from generic chatbots by being context-aware, tool-equipped, and specialized for Splunk workflows.
Navigation
- Parent: Agents
- Next: Purpose-Built Agents
- Related: Creating Agents | Chatting with Agents
What Agents Are in Deslicer
Agents in Deslicer are AI-powered assistants that you chat with. Each agent has a name, a system prompt that defines its role and behavior, and a set of attached tools. Tools let agents perform actions — run SPL, search the web, query GitHub, send Slack messages, and more. Agents use these tools when your questions require them.
How They Differ from Generic Chatbots
Generic chatbots like ChatGPT don't see your environment. They guess field names, invent sourcetypes, and produce SPL that often fails in production. Deslicer agents are different in three ways:
- Context-aware via MCP — When you connect Splunk via MCP, agents inspect real indexes, sourcetypes, fields, and configurations. They generate SPL that matches your environment.
- Tool-equipped — Agents call tools to run searches, fetch metadata, and perform actions. They don't just suggest; they can execute and show results.
- Specialized — Purpose-built agents like Search Ninja and Splunk Sensei are tuned for specific tasks. They follow best practices and explain their reasoning.
Agent Lifecycle
You create an agent, configure it, chat with it, and iterate:
- Create — Use Agent Builder or a template to create a new agent. You choose a name, model, and initial tools.
- Configure — Attach integrations (Splunk MCP, Exa, GitHub, etc.), write or refine the system prompt, and adjust model settings.
- Chat — Start conversations. The agent responds using its tools and context. You ask follow-ups, refine queries, and get actionable output.
- Iterate — Update the agent's prompt, add or remove tools, or switch models based on what works. Save changes and continue chatting.
Agents persist across sessions. You return to the same agent and conversation history whenever you need it.
