An AI agent workspace
Connecting existing agent tools with task instructions, shared notes and cloud infrastructure for research and content work.
- OpenClaw
- Hermes
- Docker
- Discord
- Obsidian
My role
I configure and operate a workspace built around OpenClaw and Hermes, define tasks and permissions, and check critical configuration. These tools provide the agent foundations; my contribution is their integration, task setup and operation.
The problem
Recurring work needs more than a single conversation. Agents need instructions, access to the right tools and a way to retain useful context. I wanted research, content work and operating notes to fit into a shared process.
How it works
A task arrives through Discord or a schedule
The assigned agent uses its instructions and tools
Shared notes provide working context
Outputs and operating notes support review and the next task
Technical decisions
Give the agents distinct responsibilities
The workspace separates technical and documentation tasks from content and research tasks. This makes it easier to specify what each agent should do and what access it needs.
Keep working knowledge outside a single chat
An Obsidian vault holds documentation and task context. Written procedures help carry useful information between sessions.
Treat hosting as part of the system
Docker, cloud hosting and protected API access are part of operating the workspace. Model providers are selected for the task, and configuration includes fallback paths where available.
Outcome and limitations
This gives recurring tasks a shared operating context and a place for the procedures learned while running them. The existing agent frameworks supply the core capabilities, and the shared instructions and notes connect them to my recurring work.
Autonomy is bounded by the tools and permissions given to an agent. Outputs and consequential actions need review. Provider changes can affect output quality and cost, so fallback paths also need checking.