Gods as agent archetypes, running inside Hermes Agent. Zero config. 30-second install. No API keys.
Olympus turns Hermes Agent into a pantheon. Each "god" is a focused, reusable skill with one clear role and a small set of tools — an agent archetype you can summon on demand. Hermes is the orchestrator (the OS); Olympus is just a convention for dividing labor among specialized agents, not a separate process.
- No Claude. No
ANTHROPIC_API_KEY. No Docker emulation. - Uses the Hermes tools you already have, plus local Ollama (only for the two schema/validation gods).
- Free: local Ollama + Hermes (MIT).
git clone https://github.com/fesenko-code/olympus.git ~/.hermes/skills/olympus
cd ~/.hermes/skills/olympus && ./setup.shThat's it. setup.sh is idempotent: it clones (or updates) the repo, symlinks the
seven god skills into your Hermes skills folder, and runs hermes profiles reload.
No prompts, no keys.
On this machine
HERMES_HOMEalready points at your brain profile, so the script resolves the correct skills path automatically — you don't need to set anything.
| God | Archetype | Role | Tools | PydanticAI |
|---|---|---|---|---|
| Zeus | Orchestrator | Plans work, delegates to other gods | delegate_task, read_file |
no |
| Athena | R&D / Architect | Structures research, emits strict schemas | search_files, write_file, patch |
yes (result_type) |
| Hermes-ag | API gateway | External calls, integration, web extract | delegate_task, web_extract |
no |
| Hades | Memory / Data Lake | Persists & recalls via TencentDB | memory_tencentdb_* |
no |
| Hera | Security / Validation | Model-validated guardrails & reviews | patch, read_file, clarify |
yes (model validation) |
| Apollo | Content | Writes structured docs & artifacts | write_file |
no |
| Hephaestus | Infrastructure | Scripts, automation, builds | terminal, execute_code |
no |
Each god is also mapped to a sociomics (Socionics) type — see references/sociomics-map.md.
After hermes profiles reload, summon a god by its olympus-<name> skill from any
Hermes session:
Use olympus-zeus to plan and delegate this feature across the other gods.
Use olympus-athena to design the data schema and emit a Pydantic model.
Use olympus-hera to validate this config against the security policy.
Use olympus-hades to remember this decision and recall it next session.
Ready-made slash-style invocations (a memorable command wins):
/olympus:zeus <goal>— break a goal into a delegated plan/olympus:athena <topic>— research + structured output/olympus:hera <artifact>— validate / review/olympus:hades <fact>— store or recall memory
- Hermes Agent = orchestrator (OS). "Olympus" is a convention for dividing work inside Hermes, not a new runtime.
- PydanticAI only inside Athena/Hera — used purely for structure and validation. Everything else is plain Hermes tool use.
- Free stack: local Ollama + Hermes (MIT). Nothing leaves your machine
unless a god's tools (e.g.
web_extract) are explicitly invoked.
olympus/
├── README.md
├── setup.sh
├── references/
│ └── sociomics-map.md
└── skills/
├── zeus/ SKILL.md
├── athena/ SKILL.md (PydanticAI)
├── hermes-ag/ SKILL.md
├── hades/ SKILL.md
├── hera/ SKILL.md (PydanticAI)
├── apollo/ SKILL.md
└── hephaestus/ SKILL.md
MIT. Original Hermes skills only — no third-party or copied code.