A local agent built to protect your code. Monitors execution, auto-heals errors in real-time, and audits security policies, strictly respecting your uneven.config.ts entirely on your machine.
curl -fsSL https://uneven.dev/install.sh | bashUneven AI is an autonomous, localized, and ultra-fast AI agent that works alongside you directly in your terminal.
While you develop, it observes execution streams in real time, detects errors, proposes and applies automatic fixes, conducts continuous security audits (Pentest Mode), executes data analysis with absolute sovereign control, and logs every action forensically — running 100% on your machine without transmitting code to the cloud.
Built from the ground up to be bank-ready, Uneven excels in modern software ecosystems (TypeScript, Python, Rust…) as well as legacy infrastructures (COBOL, Assembly, Delphi). It operates as an embedded senior developer and dedicated security specialist working silently alongside your workflows.
How Uneven AI delivers localized execution velocity while protecting your API quota budgets.
Traditional AI assistants devour massive API quotas by loading unoptimized project data. Uneven AI implements a highly advanced awareness model that processes project context with surgical precision. Complex queries that typically drain 500,000 tokens are resolved using less than 12,000, delivering highly accurate solutions in under 5 seconds while slashing output token overhead by 90% during autonomous error correction.
Deep codebase scanning and continuous security audits no longer paralyze developer workflows. Uneven AI executes ultra-fast automated processing directly on your hardware, bypassing cloud latency for heavy analytical operations. By computing intensive file analysis locally, it achieves unprecedented execution speeds while reducing total external data transmission requirements by 99.9%.
Data privacy is embedded directly into the operational core. When interfacing with local databases or performing analytical tasks, Uneven enforces strict mathematical boundaries. Production datasets and sensitive core architecture remain completely isolated from external visibility. You receive comprehensive executive intelligence and dynamic visualizations with the absolute assurance that sensitive project contents never leave your machine.
Every capability works locally, in real time, configured via a single TypeScript object.
Talk to your codebase in plain text. Ask Uneven to write complete files from scratch or perform targeted refactors. It analyzes context, drafts surgical code corrections, and saves local history in Markdown so you can review and revert changes anytime.
interactive shell · clean diffs · git-friendlyLet Uneven watch your dev server or test suites in the background. The moment a crash or compilation error triggers, it intercepts the stack trace and suggests a ready-to-apply patch instantly, cutting down debugging time to zero.
background diagnostics · auto-healing · stream watcherIndex source files, PDFs, spreadsheets, and active databases entirely offline. With native hardware acceleration (NVIDIA CUDA or Apple Silicon Metal), you run high-performance models locally, keeping your proprietary logic completely safe from cloud leaks.
offline-first · hardware native · secure bi analystKeep your workspace safe with real-time OWASP threat scanning, secrets detection, and malware verification before you commit. Runs as a headless validation pipeline to secure your code, dependencies, and infrastructure in one step.
malware & secrets scan · continuous pentesting · ci/cd checks1export default { 2 3 // brain engine 4 brain: { 5 provider: 'local', // or 'openai' | 'claude' | 'gemini' 6 model: 'llama3.2', 7 temperature: 0.3, 8 }, 9 10 // knowledge sources 11 knowledge: { 12 dirs: ['./src', './docs'], 13 files: ['./context.md'], 14 db: { 15 url: process.env.DATABASE_URL, 16 allowedTables: ['users', 'orders'] 17 }, 18 urls: ['https://docs.example.com'], 19 }, 20 21 // terminal watcher 22 watch: { 23 terminal: 'npm run dev', 24 autoFix: true, 25 confirmBeforeFix: false, 26 }, 27 28 // data security masking 29 security: { 30 mask: ['password', 'ssn'], 31 block: ['secrets'], 32 }, 33 34 // security analysis 35 pentester: { 36 enabled: true, 37 mode: 'active', 38 target: 'http://localhost:3000', 39 static: { owasp: true, secrets: true, deps: true }, 40 bruteforce: { enabled: true, detectRateLimit: true }, 41 }, 42 43 // log configuration 44 log: { 45 path: './.uneven/log.md', 46 includeDiff: true, 47 }, 48 49}
Download and install the toolkit securely via cURL. No external dependencies or API keys required.
curl -fsSL https://uneven.dev/install.sh | bashScaffolds uneven.config.ts, creates .uneven/ folder and downloads the default local model (~1.5GB, one time only).
uneven initLaunch the interactive shell to query your codebase, review commits, and run security audits.
uneven Uneven AI spawns alongside your dev process, reading stdout and stderr in real time. The moment an error appears, it diagnoses the cause, applies a surgical fix and writes a detailed entry to .uneven/log.md — with file, line, column, diff and recommendation.
Uneven's Master Series now features **Modular Security Domains**. By isolating business logic from infrastructure, we ensure that sensitive code never leaves your local machine. Every AI suggestion is audited against a defensive constitution to ensure malware and exploits never reach your source code.
⚠ Active mode should only be used on systems you own or have explicit authorization to test.
Uneven executes all file scans, process monitors, and logic routing locally on your device. Swap language models seamlessly by connecting your preferred API keys or local Ollama daemons.
Runs fully on-device. No API keys, no external network requests, and zero metadata leakage. Optimized natively for offline enterprise repositories.
provider: 'local'Superior cloud logic with ultra-low latency. Connect your Google AI keys for quick, cost-effective reasoning fallbacks.
provider: 'gemini'Integrates with any custom model installed in your local Ollama instance (such as Llama 3 or Qwen) for private offline tasks.
provider: 'ollama'Integrates GPT-4o models directly into your terminal. Combines OpenAI intelligence with local project execution boundaries.
provider: 'openai'Connects Claude Sonnet models. Ideal for deep codebase reasoning and complex multi-file logical refactoring.
provider: 'claude'Get started with Uneven AI for free and run a fully secure, localized developer companion directly inside your terminal.