A project-first Agent Operating System for teams.
Build, orchestrate, and test AI agents with unified runtime, memory, and resource registry.
Docs Site: https://monkeyhlj.github.io/HyperAgents/
HyperAgents is designed for teams that want more than a chat demo. It provides a clear project boundary, structured resources, a provider-agnostic runtime, and memory capabilities that can evolve from local development to production-grade deployment.
- Project-first model: all resources belong to a project and follow visibility policies.
- Unified resource system: Agent, Workflow, Tool, Skill, MCP, Knowledge Base.
- Runtime execution layer: route chat requests to OpenAI-compatible or local providers.
- Memory service: write/search memory, auto embedding, retry queue, semantic retrieval.
- Registry APIs: project-scoped registration and cross-project public discovery.
- Full-stack workspace: FastAPI backend and Vue 3 frontend workbench.
flowchart LR
UI[Frontend Workbench\nVue + Vite] --> API[Backend API\nFastAPI]
API --> Runtime[Runtime Executor]
Runtime --> LLM1[OpenAI-Compatible Provider]
Runtime --> LLM2[Local Provider\nOllama/vLLM]
API --> Memory[Memory Service]
Memory --> Embed[Embedding Providers]
API --> DB[(PostgreSQL + pgvector)]
API --> Registry[Resource Registry\nMCP/Tool/Skill]
- backend: FastAPI service, runtime, memory, DB models, Alembic migrations.
- frontend: Vue 3 application for project/resource/workbench operations.
- docs: bilingual node-by-node docs, quick start, and testing playbooks.
- .env.example: centralized environment template.
- Copy environment template:
copy .env.example .env- Start backend and frontend via scripts:
./scripts/start-backend.ps1 -Environment dev -RunMigrations
./scripts/start-frontend.ps1 -Environment dev -InstallFor detailed setup, see:
HyperAgents is under active iteration. APIs and docs are evolving toward a stable v1 workflow for project operations, provider integration, and memory reliability.