Description
Ax empowers developers to build robust AI applications using TypeScript by focusing on declarative signatures rather than manual prompt engineering. The framework compiles type-safe inputs and outputs into optimized LLM calls, which are then seamlessly chained into sophisticated agents, complex flows, and self-improving pipelines. This approach ensures that applications are production-ready from day one, incorporating battle-tested features like end-to-end streaming, Zod-based validation with auto-retry, and OpenTelemetry observability for deep insights.
Read More
Ax is designed with zero external dependencies, ensuring a lean and stable runtime. It supports universal deployment across Node.js, Deno, and browsers, leveraging Web Workers for sandboxed execution. A key innovation is the Multi-Objective GEPA Optimizer, which helps balance critical performance metrics such as accuracy, speed, and cost by returning Pareto frontiers. This allows developers to pick the optimal trade-off for their specific needs.
For handling complex, long-context tasks, Ax introduces the Deep Context Recursive Language Model. This feature enables long-context analysis with persistent sessions and iterative refinement, keeping extensive context out of the root prompt and significantly reducing token costs. Autonomous agents are facilitated through AxAgent, which supports ReAct loops, tool calling, child agents, context policies, and dynamic function discovery, all managed within a secure JavaScript runtime with checkpointed context management.
The framework's DSPy-style TypeScript signatures provide end-to-end type safety, ensuring that shape mismatches are caught at compile time. This is complemented by automatic prompt tuning using the GEPA optimizer, which finds the optimal instruction set for maximizing accuracy across evaluation datasets without manual prompt engineering. Ax offers a comprehensive suite of tools including AxGen for signature definition and structured output generation, AxAI for unified access to over 15 LLM providers, AxAgent for building complex agents, AxFlow for pipeline orchestration, and AxLearn for self-improving optimization.
Ax is built for production, featuring extensive test coverage, full OpenTelemetry integration for distributed tracing and detailed metrics, cost tracking per request, and enterprise-grade error handling. Its multi-runtime capability and rich type system, supporting strings, numbers, booleans, classes, arrays, JSON, images, audio, and dates, make it a versatile choice for a wide range of AI development needs. Developers can integrate with databases via GraphJin as an MCP tool, further extending the capabilities of Ax agents.
- CategoryAI Agent Builders
- DeploymentCloud
Ax's Core Features
Type-safe AI application development in TypeScript
Declarative signatures for LLM inputs and outputs
Automatic compilation of signatures into optimized LLM calls
Chaining LLM calls into agents, flows, and pipelines
Production-ready features: streaming, validation, auto-retry
OpenTelemetry integration for observability and telemetry
Universal runtime support for Node.js, Deno, and browsers
Multi-Objective GEPA Optimizer for accuracy vs. speed vs. cost trade-offs
Deep Context Recursive Language Model for long-context analysis
AxAgent for building autonomous agents with ReAct loops and tool calling
DSPy-style TypeScript signatures with compile-time validation
Automatic few-shot example tuning and prompt optimization
Unified interface for 15+ LLM providers
Built-in cost tracking and enterprise-grade error handling
Integration with databases via GraphJin MCP tools
Getting Started with Ax
Define AI capabilities: Declare inputs and outputs using type-safe signatures.
Instantiate LLM: Connect to your preferred LLM provider using a unified interface.
Build agents and flows: Chain signatures into complex agents, pipelines, or ReAct loops.
Execute and stream: Run your AI applications, with options for real-time streaming output.
Validate and retry: Leverage built-in validation and automatic retries for robust execution.
Monitor and optimize: Utilize OpenTelemetry for observability and GEPA for performance tuning.
Ax's Use Cases
- AI Agent Development
- Data Extraction & Structuring
- Content Generation
- Long-Context Analysis
- Production AI Pipelines
- Cross-LLM Application Development
- Real-time AI Applications
