Agent orchestration
WiseAI Harness
A calendar, a pipeline, and an inbox can each tell a different story about what matters this morning. I built Harness to give one agent a view across those tools and keep the final call with you.
Explore HarnessAaron Wise · Independent AI builder
I’m Aaron. I build software that gives AI agents a job and lets you see how they did it. Give my briefing agent a priority and see what lands on your desk.
Agent orchestration
A calendar, a pipeline, and an inbox can each tell a different story about what matters this morning. I built Harness to give one agent a view across those tools and keep the final call with you.
Explore HarnessLocal model evaluation
Can a local model return valid JSON and follow a word limit? WiseBench lets builders run the same tasks on their own hardware and inspect each model’s answers.
Inspect the leaderboardModel infrastructure
I built one OpenAI-compatible endpoint for a managed pool of models. Developers can connect the client they already use and follow their usage in the Storm Console.
Connect to the APIManaged agent hosting
Give your team a private agent that keeps working when your laptop is closed. I host each subscription in its own container. Your agent keeps its files, memory, and skills between sessions, so a scheduled morning brief can pick up where yesterday’s work ended.
Open the workspace in your browser or connect your own Telegram or Discord bot. Built on Hermes Agent by Nous Research.
One isolated agent with persistent storage. Model access through your agent’s OpenAI-compatible Toknado endpoint is included each month, with a fair-use limit of 2,000 requests per hour per instance.
Explore Hermes hosting and pricing ↗Managed agent hosting
Your own always-on OpenClaw workspace, hosted in a dedicated container. Use the browser panel and isolated terminal, install skills, and keep your workspace on its own persistent volume.
Explore OpenClaw Private ↗One private instance with Toknado model access, daily backups, and data export. Inference is subject to fair use.
See plan details ↗Inside my LM Studio build
When I ask my local model to prepare a briefing, it needs access to sources, memory, and somewhere to put the result. I connected those tools through 21 Model Context Protocol servers.
Explore the system mapSome of my tools start with a problem in my own workflow. Others start with a question, like whether ten agents can share the same memory.
I’m building an email workspace that takes a goal like “bring customers back” and helps choose an audience, draft a campaign, and plan its follow-ups.
I gave ten Discord agents a shared memory. The Command Center shows me which agents are working, what they did, and how many tokens each model request used.
I built this workspace to find recurring problems in Reddit communities and keep track of the ideas I choose to prototype.
I turned ten AI and security puzzles into a timed browser game. Players work through clearance gates in a fictional admin terminal, accompanied by a dial-up soundtrack.
Helping people set up OpenClaw led me to collect the agents, skills, and hands-on training labs I wanted to share with other builders.
My library includes a daily briefing skill, newsletter templates, and a workflow for turning meeting notes into assigned tasks.
About Aaron
Have a process you’d like to automate? Tell me who handles it today and what a good result looks like.
Open to opportunities in senior AI engineering, AI product, and automation.