Description
Yada is an open-source AI agent harness designed for applications built on Node.js, providing a robust framework for embedding AI agents directly into your software. This free and MIT-licensed tool allows developers to connect any large language model (LLM) to their existing HTTP APIs, SQL queries, or MCP servers, transforming them into callable tools for the AI agent.
Read More
The core functionality of Yada revolves around enabling users to interact with your application's backend through natural language chat. Whether it's placing orders, editing profiles, or sending invoices, if you define the capability, users can execute these actions by simply asking the agent. Yada ensures that all inputs are validated and permissions are checked on every API call, maintaining security and control.
One of Yada's key strengths is its flexibility in model integration. It supports popular models like Anthropic, OpenAI, Gemini, DeepSeek, and local models via Ollama and llama.cpp. Switching between models is a straightforward configuration change, eliminating the need for extensive code rewrites. The platform also offers durable storage, persisting conversations and run states in SQLite, PostgreSQL, or MySQL, which allows agents to seamlessly resume operations even after crashes or deployments.
Developers can extend their agents with custom skills and integrate a chat UI that perfectly fits their product's aesthetic. Yada provides a `yada agent harness`, `yada/client` browser SDK, and a `yada/widget` chat widget, along with comprehensive guides for every part of the harness. This includes one-command setup, reusable skills, model and provider configurations, approved actions, and deployment strategies.
Security is paramount with Yada, as it operates within your backend, using your keys and authentication mechanisms. It only accesses what you explicitly permit, with permissions checked on every call. The system also supports advanced features like approved writes, where mutating tools require human confirmation before execution, and sub-agents for nested, complex tasks. Yada's architecture is a server-side TypeScript runtime, ensuring a scalable and maintainable solution for integrating powerful AI capabilities into your applications.
- CategoryAI Agent Builders
- Pricing ModelFree
- DeploymentSelf-hosted
- PlatformsNode.js
Yada's Core Features
Open-source AI agent harness for Node.js
Connects any LLM to existing APIs (HTTP, SQL, MCP)
Supports Anthropic, OpenAI, Gemini, DeepSeek, Ollama, llama.cpp models
Enables users to perform actions via chat (search, book, apply, update)
Input validation and permission checks on every API call
Persistent conversations and run states (SQLite, PostgreSQL, MySQL)
MIT licensed
Extensible with custom skills
Includes browser SDK and chat widget
One-command setup for agents
Approved actions for write operations
Durable storage for conversations and OAuth sessions
Sub-agent support for nested tasks
Encrypted storage with AES-256-GCM
How to use Yada?
Install Yada: Set up the Yada agent harness in your Node.js application using npm.
Configure Model: Select and configure your preferred AI model provider (e.g., OpenAI, Anthropic, Ollama) by updating a simple config file.
Define Tools: Turn your existing HTTP APIs, SQL queries, or MCP servers into callable tools for the AI agent.
Integrate UI: Embed the `yada/widget` chat UI into your application or use the `yada/client` browser SDK to build a custom interface.
Set Permissions: Define rules and permissions for what the agent can access and execute within your application.
Deploy Agent: Package and deploy your Yada-powered agent, ensuring durable storage is configured for persistence.
Yada's Use Cases
- Customer Service Automation
- Internal Tooling
- E-commerce Operations
- Data Entry & Updates
- Booking & Scheduling
- Application Management
- Knowledge Retrieval
- Process Automation
