A personal AI operations team, not one general-purpose assistant.
Seven specialized AI agents that build software, review code, monitor infrastructure, manage personal operations, and teach through the same tool-enabled runtime.
Explore Hermes Agent- Named agents
- 7
- Enterprise applications reviewed
- 10
- Pull requests reviewed
- Hundreds
- Average initial review
- ~5 min
A specialist receives the work. Hermes provides the operating layer.
I built the fleet around a simple constraint: recurring work should not begin with re-explaining the job. Each named Hermes profile owns a clear operational domain, with its own instructions, memory, reusable skills, credentials, tools, and conversation history.
The agents meet work through the interface that fits it. I can direct them in Discord, a GitHub webhook can ask Rhadamanthys to review a pull request, and schedules can ask Atlas to produce an infrastructure report. They do not stop at generated advice: they inspect repositories, use browsers and APIs, modify files, run tests, verify external state, and return the result to the originating conversation.
Rhadamanthys is the clearest example of the model at scale. It has reviewed hundreds of pull requests across 10 enterprise applications, with an average initial review time of about five minutes. The same profile isolation keeps those code-review procedures separate from my personal product work, finances, study sessions, and trading education.
Seven profiles with separate context, tools, and responsibilities.
Prometheus
Personal product builder
Builds, tests, previews, and ships my own web products and portfolio.
Rhadamanthys
Autonomous code reviewer
Responds to review-request webhooks, inspects diffs, and returns cited findings to GitHub and Discord.
Atlas
Cloud operations analyst
Investigates infrastructure behavior and produces scheduled cloud-cost reporting.
Hephaestus
Enterprise software engineer
Handles focused implementation, debugging, and operational fixes across application repositories.
Themis
Executive assistant
Supports inbox workflows and answers personal-finance questions from a read-only source of truth.
Athena
Teacher and study partner
Teaches technical concepts through intuition, guided practice, and understanding checks.
Plutus
Trading mentor
Builds risk-first trading habits through market context, planning, and journal review.
The fleet works where the task begins.
Trigger
A Discord message, GitHub webhook, or schedule starts a scoped run.
Act
The assigned profile loads its own memory and skills, then uses real tools.
Verify
Tests, read-backs, and live checks ground the result before delivery.
Designed for repeated operational work.
- Seven named profiles with domain-specific responsibilities and isolated context
- Conversational, webhook-triggered, and scheduled execution paths
- Persistent memory and reusable skills that compound operating knowledge
- Real tool use across code, browsers, APIs, files, and external services
- Verification and delivery back to the conversation where work began
- Hundreds of pull requests reviewed across 10 enterprise applications