📖 What is A2E?
A2E (Agent-to-Execution) is a declarative protocol that enables AI agents to generate and execute workflows without arbitrary code execution. Inspired by A2UI and MCP protocols, A2E provides a secure, controlled environment for agent-driven task automation.
✨ Key Features
🔒 Secure
Only pre-defined operations can be executed. No arbitrary code execution.
📝 Declarative
Agents describe what to do, not how to do it.
🧠 RAG Integration
Semantic search reduces token usage by 60-80%.
⚡ Efficient
HNSW index for fast vector search. Caching and retry logic built-in.
🔐 Production Ready
Rate limiting, authentication, monitoring, and audit logging.
📚 Well Documented
Complete specification, JSON schemas, and implementation guides.
📚 Documentation
💻 Quick Example
{"type":"operationUpdate","operationId":"fetch-users","operation":{"ApiCall":{"method":"GET","url":"https://api.example.com/users","outputPath":"/workflow/users"}}}
{"type":"operationUpdate","operationId":"filter-active","operation":{"FilterData":{"inputPath":"/workflow/users","conditions":[{"field":"status","operator":"==","value":"active"}],"outputPath":"/workflow/active-users"}}}
{"type":"beginExecution","executionId":"exec-1","operationOrder":["fetch-users","filter-active"]}🔧 Installation
# Clone repository
git clone https://github.com/MauricioPerera/a2e.git
cd a2e
# Install dependencies
pip install -e .
# Start server
python server/a2e_server.py --config a2e_config.json --port 8000📦 Operations
A2E supports 8 core operations:
- ApiCall - HTTP requests
- FilterData - Array filtering
- TransformData - Data transformation
- Conditional - Conditional execution
- Loop - Array iteration
- StoreData - Persistent storage
- Wait - Execution delay
- MergeData - Data merging
🔗 Links
Repository: github.com/MauricioPerera/a2e
License: Apache 2.0
Version: 1.0.0