Project 05
AI Agent Infrastructure
An experimental AI agent environment using OpenClaw, Hermes, LLM integrations and external tools to create agents capable of researching, interacting with systems and executing technical workflows.
What it is
I run AI agents on my own servers with OpenClaw and Hermes. An agent takes a request in plain language, plans the steps and uses tools to carry them out: a shell, a browser, files, web search and APIs.
How it works
The agents use Telegram as their interface, so a request can come from a phone. Several agents can run side by side on one gateway, each with its own bot. They keep notes and reusable procedures between sessions, so a recurring task follows the same tested steps each time.
Control
Credentials are kept in a vault and filled in by the tool, never typed into a chat. Risky actions wait for approval. I treat what an agent reports as a claim and check it against real output before relying on it.
What it does
I use the agents for day to day technical work: deploying websites, editing automation workflows, investigating errors and auditing systems.
- How it starts
- A person sends an instruction or request.
- Where results go
- A person, with authorized actions and outputs delivered through connected systems.
- Stack
- AI Agents, LLM Integrations, OpenClaw, Hermes