Uneven AgentLocal code repair

Uneven Agent is a locally run AI coding agent designed to watch a development server rather than wait for chat prompts. When a runtime or compilation error occurs, it analyzes the stack trace, applies a targeted fix, and records each change in a local Markdown file. The project says file scanning and indexing happen on local hardware, with zero bytes of code sent to the cloud. Users bring their own API key and can connect Native engine, OpenAI, Anthropic, Gemini, or Ollama. The tool supports modern languages such as TypeScript, Python, Go, and Java, along with legacy languages including COBOL and Delphi. Personal use is free, while commercial projects require a license.

freemium

AI coding assistantautomated code repairlocal AI agentdeveloper toolscode privacyterminal watcherbring your own API keylegacy language support

IndexedAugust 27, 2026

0.0 (0 Number of reviews)

Try Now

Uneven Agent takes a different approach from a chat-based coding assistant. It runs as a background watcher beside the terminal, monitors a development server, and is designed to respond when the application breaks. The project says it can inspect a stack trace, apply a targeted repair, and log every change in a local Markdown file, reducing the need for developers to stop and debug manually.

A watcher built around the terminal

According to the official site, the agent reads the development server’s execution stream and reacts when it captures an error. Before writing code, it performs security checks covering issues such as SQL injection and supply-chain attacks. The tool is presented as a parallel daemon rather than a conversational assistant, although the publicly available technical details remain relatively high-level.

Local indexing and bring-your-own keys

Uneven Agent’s privacy model centers on local processing. The company says file scanning and indexing run on local hardware and that zero bytes of code are sent to the cloud. Users must provide their own API key and can choose among five providers: Native engine, OpenAI, Anthropic, Gemini, and Ollama.

When a cloud model is connected, the official description says the service sends only the prompt context needed for a specific request, rather than the entire project. A local index caches repository context, which is intended to reduce token use in later requests.

Languages, licensing, and cost

The website says the agent supports TypeScript, Python, Go, Java, and other modern stacks, as well as legacy languages such as COBOL and Delphi. Installation requires a single command, with the company claiming that users can go from installation to a first result in about two minutes.

Uneven Agent is released under the BSL 1.1 (Business Source License), and its source code is available. Personal use is free, while commercial projects require an authorized license. For developers prioritizing local indexing and code-privacy claims, the model is practical on paper; its deeper mechanisms and limits still need to be checked against the project’s GitHub source and documentation.

Method: This page was prepared with AI assistance from the official site or repository, with key numbers automatically checked against that source. Features and pricing may change; verify them at the source. official product page

Pros & Cons

Pros

Runs locally, with the company claiming that code stays on the user’s machine
Automatically repairs errors and records changes in a local Markdown file
Supports multiple AI providers and both modern and legacy programming languages
Free for personal use

Cons

Public technical details are limited, so deeper mechanisms and boundaries require checking the source and documentation
Commercial use requires a paid license

Frequently Asked Questions

Is Uneven Agent free?

The official site says personal use is free. Commercial projects require an authorized license, so teams using Uneven Agent for business work may need to arrange commercial terms. The source does not provide a specific commercial price.

Which AI providers does Uneven Agent support?

The official provider list includes five options: Native engine, OpenAI, Anthropic, Gemini, and Ollama. Uneven Agent uses a bring-your-own-key model, meaning users provide the API key needed for the provider they select.

Does Uneven Agent send code to the cloud?

The company says that zero bytes of code are sent to the cloud and that file scanning and indexing take place on local hardware. If a cloud model is connected, the official description says only the prompt context required for a particular request is sent, rather than the entire project.

Which programming languages does Uneven Agent support?

The website says Uneven Agent supports modern stacks including TypeScript, Python, Go, and Java. It also lists legacy languages such as COBOL and Delphi. The publicly supplied information does not provide a more detailed compatibility matrix or explain the limits of support for each language.

Explore More

Similar Tools

KALYPSO

KALYPSO-v1.1L is an open 14B coding model from GenomaLabs-com, published on Hugging Face and built on Qwen2.5-Coder-14B. It was fine-tuned on 18,049 curated agentic and coding examples from the Kraken dataset. The release includes model weights, the complete training dataset, and a published decontamination process for HumanEval and MBPP. That level of disclosure makes KALYPSO useful for developers researching code generation, tool-using agents, and reproducible fine-tuning. It is not presented as a universal coding benchmark winner, however: detailed technical reports and comparative evaluations are limited, while local deployment still requires meaningful GPU memory.

Locofy

Locofy is an AI design-to-code tool that turns Figma and Penpot designs into structured frontend code. It sits between visual design tools and coding agents such as Cursor and Claude Code, giving developers a UI foundation before business logic and application behavior are added. The platform supports React, React Native, HTML-CSS, Flutter, Vue, Angular, and Next.js, with workflow options including CLI and IDE integrations. Locofy also highlights SOC 2 and ISO certifications alongside cloud and on-premises deployment. It is not a replacement for frontend engineering, but it can reduce repetitive layout work and speed up the path from a polished design to a working prototype.

Template Empire

Template Empire offers production-ready Next.js full-stack templates and UI kits, uniquely distinguished by a rigorous AI-driven quality assurance process. Each release undergoes cross-auditing by 15 AI models (Claude, Codex, Gemini) and passes 23-24 quality gates, complete with a signed report. Full-stack templates include authentication, payments, and admin modules, support Docker, and are available via a one-time purchase with no subscriptions.

Polygram Coding Agent

Polygram Coding Agent is an AI programming assistant for IDEs like VS Code and Cursor. It uses a multi-agent workflow and model routing to handle the entire application development process, from product planning and UI design to code generation. Positioned as a mobile and web application builder, it's ideal for developers and teams focused on rapid prototyping.

Stellar AI

Stellar AI is an AI script generator tailored for game developers, specializing in FiveM's QBCore/ESX frameworks and the Roblox platform. It bypasses complex prompts, generating complete Lua scripts with fxmanifest from everyday English descriptions. It can also fix broken code. A free tier is available, with paid subscriptions starting from £10/month.

AuraIntel

AuraIntel is a real-time AI coding assistant designed as a desktop overlay, claiming to be undetectable during Zoom, Teams, or Discord screen shares. It captures problems, generates code, and automatically types solutions at the cursor, specifically targeting programming interviews. However, official technical details and pricing remain undisclosed, and its stealth design raises significant academic integrity concerns.

Open-source Alternatives

DeepSeek-Reasonix: Terminal AI Coding Agent for DeepSeek Models

DeepSeek-Reasonix is a terminal-based AI coding agent built for DeepSeek models, optimized to lower token costs through prefix-cache friendly prompts. Configuration is centralized in a single reasonix.toml file, and any OpenAI-compatible endpoint can be swapped in. A plugin layer allows MCP servers and Extension Protocol v1 sidecars to contribute tools and resources, while startup context injection and stale-output pruning keep long sessions lean. Written in Go, it ships as a static single binary that cross-compiles to six platforms, and also installs via npm and Home. Licensed under MIT, the repository officially claims 18,803 stars as of collection time.

code-graph-rag: Open-source RAG system leveraging knowledge graphs and LLMs for monorepos

code-graph-rag is an open-source RAG system that leverages knowledge graphs and large language models (LLMs) to navigate complex, multi-language monorepos. It enables natural language queries, deep code understanding, and editing across vast codebases, helping developers manage intricate projects more efficiently. The project is primarily written in Python and is licensed under the MIT License. As of collection time, it has 2,333 stars on GitHub.

OpenMonoAgent.ai: Terminal-Native Coding Agent Powered by Local LLMs

OpenMonoAgent.ai is an open-source, permanently free terminal-native programming agent driven by local large language models. It offers unlimited token usage and ensures privacy by running entirely on your machine. Designed for developers, it installs with a single command, eliminating subscription fees and data leakage concerns. The primary language is C# and the license is listed as Other.

WeChat Mini App Radar: AI-Powered Tech Radar for Mini Programs

WeChat Mini App Radar is an open-source, AI-powered technology radar for the WeChat mini-program ecosystem. It turns frameworks, component libraries, tools, and SDKs into filterable, comparable data, offering Radar, Compare, Advisor, Doctor, and Weekly views. Built with Next.js, TypeScript, Tailwind CSS, and Drizzle ORM, cataloging 236 resources. Licensed under GPL-3.0.

Tabby: Self-hosted AI coding assistant, an on-premises alternative to GitHub Copilot

Tabby is an open-source, self-hosted AI coding assistant designed as an on-premises alternative to GitHub Copilot. It is self-contained, requiring no external database, and offers an OpenAPI interface. Tabby runs on consumer-grade GPUs and supports RAG-based code completion and chat. Developed in Rust and licensed under Apache 2.0, it provides a fully controllable coding assistant solution.

lanhu-mcp: Open-Source MCP Server to Boost Requirement Analysis Efficiency

lanhu-mcp is an open-source Model Context Protocol (MCP) server designed for AI-driven team collaboration. It automatically parses requirement documents, generates frontend and backend code, and provides design asset downloads. Built with Python, it officially claims to improve requirement analysis efficiency by up to 200% and integrates smoothly into existing development workflows. This tool is particularly useful for accelerating prototyping and reducing manual coding effort. The project is licensed under MIT and had 1614 GitHub stars at the time of collection.

Try Now