Enaya Agent

Task-delegation AI agent with multi-agent orchestration, deep research, and structured planning.

Built on Hermes Agent architecture principles — local-first, multi-provider, extensible.

Quick Install

# From source
git clone https://github.com/tnvmac-web/Enaya-Agent.git
cd Enaya-Agent
pip install -e ".[dev]"

# Or install directly (when published)
pip install enaya-agent

Quick Start

# Interactive chat
enaya chat

# Single query
enaya chat -q "Research the latest LLM agent architectures"

# Delegate a complex task
enaya delegate "Build a REST API with authentication and tests"

# Deep research
enaya research "Current state of distributed caching" --depth deep

# Create a plan
enaya plan "Implement user authentication system" --complexity moderate

# Switch model
enaya model openrouter:anthropic/claude-3.5-sonnet

# Setup wizard
enaya setup

Core Capabilities

Architecture

┌─────────────────────────────────────────────────────────────┐
│                     Enaya Agent                             │
├─────────────────────────────────────────────────────────────┤
│  AIAgent (run_agent.py) — Single class, all entry points   │
├─────────────────────────────────────────────────────────────┤
│  Agent Loop          │  Prompt System      │  Provider Res  │
│  conversation_loop   │  prompt_builder     │  runtime_      │
│  context_compressor  │  anthropic_adapter  │  provider      │
├─────────────────────────────────────────────────────────────┤
│  Tools (50+)         │  Delegation         │  Memory        │
│  registry            │  orchestrator       │  5-layer       │
│  file/web/delegation │  task_queue         │  SQLite+Chroma │
│  research/planning   │  result_aggregator  │  project_index │
│  synthesis           │  policies           │                │
├─────────────────────────────────────────────────────────────┤
│  Interfaces: CLI • ACP (VS Code/Zed) • API Server • Gateway │
└─────────────────────────────────────────────────────────────┘

Configuration

# ~/.enaya/config.yaml
model: "openrouter:anthropic/claude-3.5-sonnet"
provider: "openrouter"
max_turns: 500
temperature: 0.7
toolsets: ["core", "research", "planning", "delegation", "synthesis"]
fallback_providers: []
compression_threshold: 0.50
prompt_caching: true

# ~/.enaya/.env
OPENROUTER_API_KEY=sk-...
ANTHROPIC_TOKEN=sk-ant-...
OPENAI_API_KEY=sk-...
NVIDIA_API_KEY=...

Providers Supported

Skills (Bundled)

Programmatic Integration

ACP (VS Code, Zed, JetBrains)

enaya acp  # Starts JSON-RPC stdio server

API Server (OpenAI-compatible)

enaya api-server  # Starts HTTP + SSE server

Python Embedding

from enaya.run_agent import create_agent

agent = create_agent(model="openrouter:anthropic/claude-3.5-sonnet")
result = agent.run_conversation("Research quantum computing advances")

Development

# Setup
python -m venv venv
source venv/bin/activate
pip install -e ".[dev]"

# Test
pytest tests/ -n0 -q

# Lint
ruff check src/

# Type check
mypy src/enaya/

# Coverage
pytest --cov=enaya --cov-fail-under=60

Project Structure

enaya-agent/
├── src/enaya/              # Core package
│   ├── agent/              # Agent loop & prompts
│   ├── cli/                # CLI commands
│   ├── tools/              # 50+ built-in tools
│   ├── delegation/         # Multi-agent orchestration
│   ├── research/           # Research pipeline
│   ├── planning/           # Planning engine
│   ├── memory/             # 5-layer memory
│   ├── gateway/            # Messaging gateway
│   ├── acp_adapter/        # ACP server
│   └── plugins/            # Plugin system
├── skills/                 # Bundled skills
├── tests/                  # Unit/integration/e2e
├── docs/                   # Documentation
├── pyproject.toml
├── AGENTS.md               # Development guide
├── HERMES.md               # Project context
└── README.md

Design Principles

• Platform-Agnostic Core: One AIAgent serves CLI, gateway, ACP, API server
• Observable Execution: Every tool call visible via callbacks
• Interruptible: API calls and tools cancellable mid-flight
• Loose Coupling: Optional subsystems use registry patterns
• Profile Isolation: Multiple concurrent instances via enaya -p <name>

License

MIT License — see LICENSE file.

References

• Hermes Agent Architecture: https://hermes-agent.nousresearch.com/docs/developer-guide/architecture
• Original Hermes: https://github.com/NousResearch/hermes-agent