Can Edict be integrated into different environments?

Yes, Edict offers various integration options including a Model Context Protocol (MCP) server for direct agent tool calls, Docker support for easy deployment, and full browser compatibility for client-side compilation and execution.

What role does JSON AST play in Edict?

JSON AST is Edict's canonical program format. This design choice eliminates the need for lexers or parsers, allowing AI agents to directly generate structured code that the compiler understands and processes without ambiguity.

How does Edict help AI agents write reliable code?

Edict provides a structured compilation pipeline with comprehensive checks: schema validation, name resolution, static type checking, effect tracking, and Z3-powered contract verification. It returns rich, structured errors with context, enabling agents to self-repair effectively.

What are the security benefits of using Edict?

Edict compiles to WebAssembly (WASM), providing a sandboxed execution environment with no ambient authority. Its effect system declares capabilities at compile-time, and runtime enforcement ensures secure execution for agent-generated code.

Edict

Name: Edict
Author: Sowiedu

bySowiedu

0•

開発者ツール

セキュリティとテスト

データサイエンスとML

Enables AI agents to write, verify, and execute programs by producing Abstract Syntax Trees directly as JSON.

Edict is a purpose-built programming language designed for AI agents, where the canonical program format is a JSON AST instead of traditional text-based syntax. This eliminates the need for parsing and ambiguity, allowing agents to directly generate structured programs. It features a robust pipeline including schema validation, name resolution, static typing, effect tracking, and contract verification using Z3/SMT solvers. Programs are compiled to WebAssembly (WASM) for secure, sandboxed execution, offering strong isolation and explicit capability declarations. All compiler tools are exposed via the Model Context Protocol (MCP) for seamless integration with AI agents, facilitating self-repair through structured error feedback.

主な機能

01JSON AST for direct agent program generation without parsers

02Statically-typed with refinement types and comprehensive type system

03Compile-time contract verification using Z3/SMT with counterexamples

040 GitHub stars

05Structured errors with rich context for agent self-repair

06Effect tracking for explicit declaration and verification of side effects

ユースケース

01AI agents writing, verifying, and executing programs

02Developing agent-generated code with formal verification and secure execution

03In-browser compilation and execution of AI-generated programs