Skip to main content
Aden Agent Framework aims to help developers build outcome-oriented, self-adaptive agents. This roadmap outlines our vision and progress across core architecture, tooling ecosystem, memory management, evaluation systems, and deployment infrastructure.

​
Architecture Overview

​
Core Architecture & Swarm Primitives

​
Node-Based Architecture

Implement the core execution engine where every Agent operates as an isolated, asynchronous graph of nodes.

Core Node Implementation

  • ✅ NodeProtocol with JSON parsing utilities (graph/node.py)
  • ✅ EventLoopNode with LLM conversation management (graph/event_loop_node.py)
  • ✅ Flexible input/output keys with nullable output handling
  • ✅ Node wrapper SDK for agent creation
  • ✅ Tool access layer with MCP integration
  • ✅ Graph traversal execution (graph/executor.py)
  • ✅ Node transition management
  • ✅ Error handling and output mapping
  • ✅ ExecutionResult with success/error status
  • ✅ SharedState manager (runtime/shared_state.py)
  • ✅ Session-based storage (storage/session_store.py)
  • ✅ Isolation levels: ISOLATED, SHARED, SYNCHRONIZED
  • ⏳ Performance metrics collection
  • ⏳ Resource usage tracking
  • ⏳ Health check endpoints

​
Judge in Event Loop

A separate LLM-powered judge to determine if workers finish their job.
  • ✅ Conversation Judge (Level 2) - Evaluates node completion against success criteria
  • ✅ Test Evaluation Judge - Provider-agnostic evaluation
  • ⏳ Multi-Level Judgment Integration - Automatic retry logic

​
Swarm Hierarchy

  • ✅ Judge Bee (Evaluator) - Evaluation criteria framework with confidence scores
  • ✅ Hive Coder Agent (Builder) - Forever-alive event loop with tool discovery
  • ⏳ Queen Bee (Orchestrator) - Multi-agent coordination layer
  • ⏳ Worker Bee (Executor) - Worker taxonomy and templates

​
Coding Agent Workflows

  • ✅ Goal Creation Session - Goal schema with success criteria and constraints
  • ✅ Agent Creation Flow - Template reading and dynamic tool binding
  • ⏳ Worker Agent Dynamic Creation - Runtime worker instantiation

​
Security Layer

Credential Store

✅ Multi-backend storage (encrypted file, env vars, HashiCorp Vault)✅ Template resolution with caching✅ Thread-safe operations

OAuth2 Providers

✅ Base provider pattern✅ HubSpot integration✅ Lifecycle management

Aden Sync Provider

✅ OAuth2 token sync from Aden server✅ Local fallback support

Enterprise Secret Managers

⏳ AWS Secrets Manager⏳ Azure Key Vault⏳ Audit logging

​
Tooling Ecosystem & General Compute

​
Sub-agents Parallel Execution

  • ✅ Multi-Graph Sessions - Load multiple agent graphs in single session
  • ✅ Concurrent Execution Management - Max concurrent executions config
  • ⏳ Sub-agent Execution Environment - Isolated runtime with repeatability guarantees

​
Browser Use Node

  • ✅ Web Scraping with Playwright - Headless Chromium with stealth mode
  • ✅ Browser Launch Utilities - Platform-specific browser opening
  • ⏳ Full Browser Use Node - Multi-page automation with vision-guided interactions

​
Core Graph Framework Infra

  • ✅ Node Validation - Pydantic-based schema enforcement
  • ✅ Human-in-the-Loop (HITL) - Pause/approve workflows with HITLRequest protocol
  • ✅ TUI Integration - Chat REPL with streaming support
  • ✅ Node Lifecycle Management - Start/stop/pause/resume with state persistence

​
Infrastructure Tools

  • ✅ File Operations (36+ tools) - read, write, edit, search, diff/patch
  • ✅ Web Tools - Search, scraper, Exa, News, SerpAPI
  • ✅ Data Tools - CSV, Excel, PDF, Vision, Time
  • ✅ Communication Tools (8 tools) - Email, Gmail, Slack, Discord, Telegram
  • ✅ CRM/API Integrations (5+ tools) - HubSpot, GitHub, Apollo, BigQuery
  • ✅ Security/Scanning Tools (5 tools) - DNS, SSL/TLS, Port Scanner
  • ✅ Runtime & Logging - Runtime Log Tool with L1/L2/L3 levels

​
Memory, Storage & File System Capabilities

​
Memory Tools

  • ✅ Short-Term Memory (STM) - SharedState manager for in-memory state
  • ✅ Conversation Memory - NodeConversation tracks message history
  • ⏳ Long-Term Memory (LTM) - Semantic indexing and RLM implementation

​
Durable Scratchpad

  • ✅ Filesystem as Scratchpad - File-based persistence layer
  • ✅ Checkpoint System - Save/restore execution state with TTL cleanup
  • ⏳ Message Model & Session Management - Per-message file persistence

​
Memory Isolation

  • ✅ Session Isolation - Session-local memory with data bleed prevention
  • ✅ State Management - SharedState with thread-safe operations
  • ⏳ Context Management - Message streaming and optimization

​
Eval System, DX, & Open Source Guardrails

​
Eval System

Completed Features

  • ✅ Multi-Level Evaluation (Level 0, 1, 2)
  • ✅ Goal-Based Constraints (hard/soft)
  • ✅ Success Criteria Definition with weighted metrics
  • ✅ Test Framework with LLM-based judgment
  • ⏳ Failure Recording mechanism
  • ⏳ Custom Failure Conditions SDK

​
Guardrails SDK

  • ✅ Goal Constraints - Hard/soft constraint definitions
  • ⏳ Deterministic Guardrails SDK - In-node guardrail implementation
  • ⏳ Monitoring & Tracking - Audit logs and compliance reporting

​
DevTools CLI

  • ✅ Main CLI - Run, info, validate, list commands
  • ✅ Testing CLI - test-run, test-debug, test-list, test-stats
  • ✅ TUI (Terminal UI) - Interactive chat with streaming
  • ⏳ Memory Management CLI - Memory inspection and cleanup
  • ⏳ Credential Store CLI - Interactive credential editing

​
Observability

  • ✅ Runtime Logging - L1/L2/L3 logging levels
  • ✅ Event Bus Monitoring - Real-time event streaming
  • ⏳ Log Analysis Tools - User-driven log analysis
  • ⏳ Token Tracking - Reasoning and cache token tracking

​
Deployment, CI/CD & Community Templates

​
Self-Deployment

  • ✅ Docker Support - Python 3.11-slim with Playwright
  • ✅ Agent Runtime - Multiple entry points (webhook, timer, event, API)
  • ✅ Async Entry Points - Timer, webhook, and event triggers
  • ⏳ Headless Backend Enhancements - Standardized APIs

​
Lifecycle APIs

  • ✅ Webhook Server - FastAPI-based webhook server
  • ✅ Graph Lifecycle Management - Load/unload/start/restart
  • ⏳ REST API Endpoints - Start, stop, pause, resume endpoints
  • ⏳ WebSocket API - Real-time event streaming to clients

​
CI/CD Pipelines

  • ✅ Test Execution - Test framework with pytest integration
  • ⏳ Automated Testing Pipeline - CI integration with GitHub Actions
  • ⏳ Version Control - Agent versioning system
  • ⏳ Deployment Automation - CD pipeline with rollback

​
Distribution

  • ⏳ Package Distribution - Official PyPI package and Docker Hub image
  • ⏳ Community Channels - Discord setup and contribution guidelines
  • ⏳ Cloud Deployment - AWS Lambda, GCP, Azure Functions support

​
Example Agents

  • ✅ Hive Coder Agent - Agent builder template
  • ⏳ Sales & Marketing Agents - GTM, lead generation, campaigns
  • ⏳ Analytics & Insights Agents - Data analysis, reporting
  • ⏳ Training & Education Agents - Onboarding, content creation
  • ⏳ Automation & Forms Agents - Smart entry, workflow automation

​
Open Hive

​
Local API Gateway

  • ✅ MCP Server Foundation - FastMCP server on port 4001
  • ✅ Event Bus Architecture - Real-time streaming with graph-scoped routing
  • ⏳ Local API Gateway - Lightweight server for browser access
  • ⏳ Memory Layer API - Memory read/write endpoints

​
Visual Graph Explorer

  • ⏳ Graph Visualization - React Flow integration
  • ⏳ Interactive Features - Drag-and-drop canvas
  • ⏳ Integration with Runtime - Live execution visualization

​
TUI to GUI Upgrade

  • ✅ TUI Foundation - Terminal chat with streaming
  • ⏳ Web Application - Modern web UI framework
  • ⏳ Chat Interface - Browser-based chat UI
  • ⏳ TUI Feature Parity - All commands in GUI

​
Memory & State Inspector

  • ✅ Runtime Logs Tool - Inspect agent session logs
  • ⏳ Memory Inspector UI - Shared memory visualization
  • ⏳ Node State Inspection - Click-to-inspect functionality
  • ⏳ Debug Tools - Memory diff viewer

​
Local Control Panel

  • ✅ Credential Management Backend - CredentialStore with multiple backends
  • ⏳ Credential Management Dashboard - Web UI for credential editing
  • ⏳ Swarm Lifecycle Management - Start/stop controls in browser
  • ⏳ Monitoring Dashboard - Active agents and resource usage

​
Local Model Integration

  • ✅ LLM Integration Layer - Provider-agnostic via LiteLLM
  • ⏳ Local Model Support - Ollama integration and configuration
  • ⏳ Offline Mode - Full offline functionality
  • ⏳ Model Configuration - Easy model switching in UI

​
Cross-Platform Support

  • ⏳ JavaScript/TypeScript SDK - npm package distribution
  • ⏳ Platform Compatibility - Windows, macOS, Linux optimization

​
Coding Agent Integration

  • ⏳ IDE Integrations - Claude Code, Cursor, Opencode, Antigravity, Codex CLI

​
Legend

  • ✅ Completed - Feature is implemented and shipped
  • ⏳ Planned - Feature is on the roadmap

​
Contributing

We welcome contributions! If you’d like to help build any of these features, check out our contributing guide or join our Discord community.

Build docs developers (and LLMs) love

Get started for freeTalk to us