YaoApp/yao
Yao
Yao is an LLM agent framework and low-code web app builder designed for orchestrating autonomous AI agents. It provides a platform to design, deploy, and coordinate agents with specialized personas that can plan tasks, utilize external tools, and execute multi-stage pipelines.
The project distinguishes itself through a Model Context Protocol server for connecting assistants to external binaries and HTTP services, and a gRPC remote execution engine that allows agents to manage remote servers and devices. It includes a model-agnostic provider bridge that supports dynamic switching between various AI model providers via an OpenAI-compatible API layer.
The platform covers a broad capability surface, including secure code sandboxing using Docker, remote infrastructure management with VNC streaming, and low-code development for generating administrative interfaces from JSON configurations. It also features autonomous workflow automation with recurring job scheduling and programmatic integration with messaging platforms like Discord and Telegram.
The entire system is distributed as a single-binary runtime.
Features
- MCP Server Connections - Integrates external tools from HTTP endpoints and local binaries into assistants using the standardized Model Context Protocol.
- Autonomous Agent Orchestration - Provides a framework for deploying modular AI agents that plan tasks and coordinate specialized personas.
- Provider-Agnostic Model Interfaces - Standardizes interactions with various AI services through an OpenAI-compatible API layer for dynamic model switching.
- Agent Deployment - Provides systems for provisioning and configuring isolated AI agent instances and specialized personas.
- Agentic Pipeline Automations - Deploys autonomous AI workers to execute complex multi-stage pipelines triggered by external events.
- AI Agent Orchestrators - Organizes and sequences a team of specialized agents to handle planning, research, and validation phases.
- Persistent Cloud Workspaces - Maintains long-running AI agent sessions in isolated containers that persist across inactivity.
- OpenAI-Compatible APIs - Links to any OpenAI-compatible API or local model instance to power agent intelligence.
- AI Agent Development - Provides tools for creating specialized AI assistants based on natural language descriptions and external skill integration.
- Cloud-Hosted Agent Sandboxes - Maintains persistent Linux environments with terminal and network access across multiple AI agent sessions.
- Autonomous AI Workflows - Implements autonomous pipelines that handle the full cycle of analysis, execution, and delivery for AI-driven tasks.
- AI Execution Sandboxes - Provides a secure containerized environment specifically tailored for the execution of AI agents.
- Dynamic Model Switching - Changes the active language model during a conversation to leverage different reasoning strengths or costs.
- External AI Model Connectors - Provides connectors for linking to AI providers using API hosts, keys, and model identifiers.
- Agent Persona Configurations - Defines specific agent identities and operating modes for scheduled or on-demand autonomous execution.
- AI Provider Integrations - Integrates external AI models via custom API endpoints and keys for OpenAI-compatible providers.
- Autonomous AI Agent Frameworks - Provides a comprehensive framework for building autonomous agents with planning and tool-use capabilities.
- Autonomous Workflow Automation - Provides a system for designing and executing autonomous, multi-stage AI agent workflows with recurring scheduling.
- General Purpose Task Execution - Enables agents to perform a wide array of tasks including coding and research via flexible automated workflows.
- Agentic Goal Decomposition - Uses LLMs to recursively break down high-level strategic goals into a sequence of executable sub-tasks.
- AI Agent Orchestrators - Coordinates model providers, tool registration, and the assignment of experts to robots for specialized execution.
- Agentic Workflow Executions - Processes complex goals through sequential briefing, execution, and validation stages to produce final results.
- Isolated Runtime Provisioning - Provisions temporary Docker containers with a full software stack to create isolated runtimes for individual tasks.
- Workspace Isolation - Maintains separate persistent sandboxes to keep files and conversation contexts isolated across different projects.
- Agentic Pipeline Orchestration - Decomposes high-level goals into a sequence of briefing, execution, and validation stages managed by specialist agents.
- Hybrid Execution Bridges - Bridges the local engine to a variety of remote machines, containers, or virtual machines.
- Sandboxing Environments - Proxies container runtimes to provision isolated environments for the execution of automated AI tasks.
- gRPC Tunnelling Services - Uses gRPC tunnels to establish encrypted communication with remote devices, bypassing firewalls without port forwarding.
- Messaging Platform Integrations - Provides adapters to connect autonomous AI agents to external social messaging services like Discord and Telegram.
- AI Agent APIs - Exposes OpenAI-compatible endpoints for each robot to enable programmatic triggering from external SDKs.
- gRPC Remote Execution Bridges - Ships a gRPC-based engine that bridges AI agents to remote servers for system administration.
- Remote Agent Task Execution - Connects multiple servers to a central engine to execute AI agent tasks on remote hardware.
- Code Sandboxing Environments - Runs AI agents and generated scripts within isolated Docker containers to ensure host security.
- Agent Execution Environments - Runs agent tasks within isolated Docker containers to ensure secure and controlled tool operations.
- Remote Sandbox Isolation - Secures data privacy by running agent sessions within isolated Docker containers and storing synced files locally.
- Container Isolation - Runs AI agents and browser tasks in isolated Docker containers to secure the host and maintain clean environments.
- Flow-Based Programming Models - Processes data streams using a flow-based programming model and custom extensions for business logic.
- Objective Prioritization - Transforms high-level user intent into prioritized objectives with specific success criteria for agent execution.
- Container Host Connections - Establishes connections between a central engine and a distributed network of laptops, VMs, and containers.
- Remote Agent Hardware Linking - Connects remote machines and servers to allow autonomous agents to operate on external hardware.
- Remote Command Execution - Routes natural language requests to execute OS commands and manage settings on remote servers.
- Remote Pairing Controllers - Links remote systems to a central management instance for the execution of AI agent tasks.
- Behavioral Agent Generation - Generates assistants using natural language or DSLs to define complex autonomous behaviors.
- Messaging Platform Connectors - Binds autonomous agents to messaging platforms to enable direct user interaction via chat interfaces.
- JSON-Driven Application Scaffolding - Generates functional web services and admin interfaces by translating JSON configuration schemas into code and layouts.
- Execution Control Flows - Manages concurrency limits and queue sizes to control the execution flow of agent processing stages.
- Local Agent Deployments - Runs domain-specific AI experts on local hardware to ensure data privacy and security.
- Agent Skill Integrations - Integrates external tools and APIs from a library of audited solutions to provide agents with specialized capabilities.
- AI Integration APIs - Provides programming interfaces that allow tools to incorporate chat, translation, and summarization via a built-in AI API.
- Model Context Protocol Integrations - Implements the Model Context Protocol to connect AI assistants to external binaries and HTTP services.
- Autonomous Web Research Agents - Provides autonomous systems that crawl, synthesize, and summarize information from online sources into structured reports.
- MCP Protocol Integrations - Connects agents to external services like calendars and databases using the standardized Model Context Protocol.
- Agent Workflow Customizers - Allows replacing default task agents with custom versions to modify execution behavior within AI pipelines.
- Persona Deployment - Deploys pre-tuned AI personas to handle specialized domain-expert requests such as coding and research.
- Natural Language Querying - Retrieves specific project information using natural language processing to generate Markdown reports.
- Professional Domain Modeling - Deploys specialized AI models tailored for professional industries to ensure high-quality, expert outputs.
- Model Context Protocol Servers - Implements a Model Context Protocol server to expose internal tools and data to AI models.
- Model Role Mappings - Maps AI models to specific roles like vision or audio to optimize performance based on task complexity.
- Web Search Integrations - Enables AI agents to query real-time web data via external APIs to provide context-aware responses.
- Data Analysis Agents - Includes agents designed to process structured files like CSV and Excel to generate statistical charts and insights.
- Recurring Task Automation - Executes autonomous workflows on a fixed schedule or via automatic triggers to handle repetitive operations.
- Remote Workspace Sync - Tracks and transfers incremental changes between local directories and remote persistent sandboxes using state manifests.
- Workspace Synchronizers - Transfers changed data incrementally between local and remote environments using state manifests.
- Web Content Fetching - Retrieves data, prices, and pages from external websites using a built-in scraping service optimized for LLM use.
- Web Search APIs - Retrieves web content and performs online searches using MCP-based server tools.
- Generative Admin Panels - Generates management backgrounds and internal enterprise systems using interface layouts described in JSON.
- Application Lifecycle Management - Writes code and manages the development lifecycle to create websites from natural language descriptions.
- Goal Confirmation Interfaces - Implements a streaming conversation interface to clarify and finalize objectives with users before task execution.
- Manual Task Triggering - Provides a dashboard for users to manually trigger and confirm the goals of automated processes.
- Remote File Synchronization - Synchronizes project files between local machines and remote AI agent workspaces.
- Web Scraping - Extracts full content from URLs to gather detailed data from websites for agent use.
- Webhook Triggers - Exposes HTTP endpoints that immediately initiate autonomous agent workflows upon receiving external requests.
- Autonomous Robot Triggers - Starts autonomous workflows using fixed intervals, background monitoring, or external HTTP webhooks.
- Agent Tool Prioritization - Enables customization of pipeline stages to control tool prioritization and output formatting during agent execution.
- Recurring Job Scheduling - Schedules tasks to run at fixed times or regular intervals based on specific timezones.
- Containerized Tool Bridges - Exposes tool servers running inside containers to agents via gRPC for extended capabilities.
- Single-Binary Deployments - Bundles the entire execution engine, server tools, and interfaces into a single executable for simplified deployment.
- SSH Deployment Tools - Uses SSH to automate the pushing of artifacts and configuration of Docker and HTTPS on remote servers.
- Firewall Traversal Mechanisms - Establishes gRPC tunnels to reach and manage machines behind firewalls without requiring port forwarding.
- Multi-Protocol Remote Node Connectors - Transforms remote servers into nodes for automated command execution and file manipulation.
- Remote Desktop Clients - Provides a live VNC view of a remote screen for visual interaction with the device.
- Remote Desktop Session Management - Establishes and manages VNC sessions to interact with the graphical user interface of remote devices.
- Remote Desktop Streaming - Streams a remote device screen to a browser for manual mouse and keyboard control.
- Connection Authentication Mechanisms - Implements secure connection authentication between remote agents and servers using OAuth 2.0.
- Command Restrictions - Controls remote agent access to OS commands through configurable whitelist and blacklist policies.
- gRPC and HTTP TLS Securings - Secures communication between the central engine and remote devices using gRPC channels and TLS encryption.
- Quality Validation Rules - Evaluates agent outputs against semantic rules to provide quality scores and verification verdicts.
- Natural Language Project Scaffolding - Automates the generation of project records and directory structures from natural language instructions.
- Project Organization - Organizes web projects using a unified index and directory-based storage for better maintainability.
- Immediate Task Executions - Implements a mechanism to start agent tasks immediately by bypassing the goal-confirmation phase.
- JSON-Defined Interfaces - Generates internal enterprise systems through JSON-described layouts for common interactions.
- Device Connection Monitoring - Tracks the online status and hardware architecture of registered remote devices in real time.
- Low-Code Generators - Generates administrative interfaces and web applications automatically from JSON configurations and natural language.
- Remote Desktop Monitoring - Provides a real-time VNC interface to visually monitor activity within remote execution environments.
- Remote File Managers - Organizes and browses file synchronization areas on connected remote devices via a management interface.
- Remote Infrastructure Management - Enables centralized control of remote servers and devices using gRPC and VNC for administration.
- Agent Behavior Simulation - Verifies agent behavior through automated tests and simulations executed within a virtual desktop environment.
- Application Deployment - Scaffolds websites and deploys them to servers over SSH using a conversational interface.
- Configuration-Driven APIs - Defines database models and API interfaces using JSON configurations to accelerate enterprise system creation.
- JSON-Based Service Scaffolding - Creates database models and API services using a JSON-based description language and flow-based programming.
- Natural Language App Generators - Creates functional utility tools like calculators from plain language descriptions for live preview.
- Business Application Frameworks - Application engine for creating models and APIs via JSON.
- Low-Code and Visual Tools - Low-code engine for building web services and applications.
