Run an AI-native company from one desktop operating layer.
Hermes Agencies helps founders and operators set up a company, define AI employees, manage projects, and route work through Hermes without running the business from prompts, terminals, and scattered task threads. Inside Builder Collective, it also serves as a concrete case study for how serious AI operating systems can be designed.
A company control layer for Hermes
Hermes is the runtime. Hermes Agencies is the operating layer around it.
Hermes Agencies does not replace Hermes. Hermes still handles execution, tools, skills, memory, sessions, and delegation primitives. Hermes Agencies adds the desktop surfaces that make multi-agent work easier to run: company structure, employees, projects, inbox, approvals, and sync.
Company workspace
Start with a company, not a blank chat.
Set the mission, define how the company runs, track active projects, and keep approvals in view from one workspace. Hermes Agencies gives operators a clear home base instead of scattering work across chats, terminals, and loose sessions.
- Single-company, founder-controlled MVP
- Mission, projects, approvals, and team structure in one place
- Local-first desktop workflow
AI employees
Give every AI employee a real job.
Employees are more than saved prompts. Each one has a role, department, identity, focus, runtime defaults, reporting line, skill set, and Hermes profile mapping. That makes the org chart usable, not cosmetic.
- Role, department, and reporting line
- Identity, focus, and runtime profile
- Base skills, extra skills, and disabled skills
- Provider and model defaults per employee
Project orchestration
Keep the navigation calm. Put the depth inside the project.
Hermes Agencies keeps the top-level interface simple, then gives each project its own focused workspace. Executive agents can hand work to the right specialists while the operator still sees owner, status, and recent movement at a glance.
- Project rail and detail workspace
- Overview, board, timeline, and outputs views
- Task detail with updates and context
- Clear ownership across executive and specialist roles
Inbox and approvals
One attention layer for what matters now.
Instead of turning every decision into another thread, Hermes Agencies uses a central inbox for approvals, blockers, review items, and executive updates. Founders get one place to see what needs attention across the company.
- Approval requests
- Blocked work
- Review-ready outputs
- Executive updates
Hermes sync
Configure in the app. Sync into Hermes.
Hermes Agencies turns visual employee setup into Hermes-ready runtime files. The app can generate employee packages, preview payloads, and sync those definitions into Hermes-managed locations.
- IDENTITY.md
- SKILL.md
- runtime.yaml
- skill-packs.yaml
- memory-layer.yaml
- hermes-profile.json
Built for company orchestration, not prompt juggling
More structured than a chatbot workspace. Easier to run than raw runtime-only workflows.
- Fast for back-and-forth work
- Thin company structure
- Harder to track ownership and delegation
- Powerful execution runtime
- Flexible, scriptable, and extensible
- More technical for day-to-day company operations
- Company is the root object
- AI employees with roles and reporting lines
- Projects, inbox, approvals, and sync in one desktop layer
Open-source app. A flagship Builder Collective case study.
Hermes Agencies can serve as a strong open-source entry point into the broader Builder Collective ecosystem. It gives members a real product to study while they learn how agent runtimes, AI team design, and operating layers fit together in practice.
Learn the runtime, the operating model, and the product workflow in one place.
Explore Hermes Agencies as the company layer, then use Builder Collective to understand how these systems are configured, reviewed, and adapted to real workflows.