YOUR CODE BREAKS.
IT FIXES ITSELF.
AI coding agent with a background watcher that catches and auto-fixes errors.
Uneven watches your dev server in real time. The moment a crash happens, it diagnoses the stack trace, applies a fix, and logs exactly what changed — so you never stop to debug.
curl -fsSL https://uneven.dev/install.sh | bashSecurity checks built-in
From SQL injection to supply chain attacks
AI providers supported
Native engine, OpenAI, Anthropic, Gemini, Ollama
Bytes sent to cloud
All file scans and indexes run on your hardware
From install to first result
One command. No configuration required.
A WATCHER,
NOT A CHATBOT.
Uneven Agent runs alongside your terminal, not instead of it.
It reads your dev server's execution stream in real time, catches errors as they happen, applies a targeted fix, and enforces security checks before code is written — logging every action to a local Markdown file so you always know exactly what changed and why.
The full client — watcher, guardrails, and terminal shell — is review the Uneven Agent source on GitHub. Supports modern stacks (TypeScript, Python, Go, Java) and legacy ones (COBOL, Delphi). Free for personal use, licensed for commercial projects.
LOCAL WHERE IT'S FASTER.
CLOUD ONLY WHEN YOU CHOOSE.
What actually runs on your machine vs. what gets sent to your model provider — and why it matters for speed and cost.
CONTEXT CACHING
Most AI assistants re-upload your whole project on every query. Uneven indexes your codebase once, keeps it warm in provider cache, and sends only the diff context each request needs — so complex queries cost less and resolve faster.
ON-DEVICE SCANNING
Codebase mapping runs on your hardware, not a remote server. There is no round trip to index your files — workspace analysis starts as soon as the scan finishes locally, independent of network latency.
LOCAL-ONLY INDEXES
Your file index, database schema, and query history are stored only on your machine. When you connect a cloud model (OpenAI, Anthropic, Gemini), only the specific prompt context for that request is sent — never your full project or database contents.
WHAT UNEVEN AGENT
CAN DO
Every capability runs locally and automatically — from chat and auto-fix to security scanning.
AI Chat for Your Codebase
Ask questions about your project in plain text, generate new files from scratch, or request targeted refactors. Uneven analyzes your codebase context, applies the edit, and saves a local Markdown history so every change is reviewable and reversible.
codebase chat · file generation · refactoring · git-friendly diffsAutomatic Error Fixing (Self-Healing Watcher)
Run Uneven alongside your dev server or test suite. When a crash or compilation error appears, it reads the stack trace, diagnoses the cause, and prepares a ready-to-apply patch — no need to stop and debug manually.
background error watcher · auto-fix crashes · stack trace diagnosisLocal Codebase & Database Indexing
Index source files, PDFs, spreadsheets, and live databases entirely on your machine. Query any of it in plain English while proprietary code and data stay off the network.
local database query · offline code indexing · private codebase searchSecurity Scanning Before You Commit
Uneven scans for OWASP vulnerabilities, hardcoded secrets, and malicious dependency behavior before code ships. Runs as a headless step in CI/CD to catch issues automatically on every push.
OWASP scanner · secrets detection · CI/CD security checks WHAT YOU CAN
DO RIGHT NOW
Four things developers use Uneven Agent for every single day.
Instant Commit Review
Ask Uneven to analyze your last commit. It reads the diff, flags vulnerabilities, spots regressions, and gives you a clear summary — before you push.
review my last commitBackground Error Watcher
Start your dev server and let Uneven watch in the background. The moment a crash happens, it diagnoses the stack trace and prepares a ready-to-apply fix — no action from you needed.
uneven watch --cmd "npm run dev"Database in Plain English
Query your database by just asking a question. Uneven translates it to SQL, runs it read-only, and returns the answer. Your data never leaves your machine.
How many orders were placed this week?Security Audit Before Deploy
Open the Security Guardian from chat with one command. Run static scans, check for secrets, and probe your staging endpoint — all before anything reaches production.
/pentest
UP IN
2 STEPS
INSTALL
Download and install the toolkit securely via cURL. No external dependencies or complicated setup required.
curl -fsSL https://uneven.dev/install.sh | bashRUN
Launch the interactive shell to query your codebase, review commits, and run security audits instantly.
uneven// System requirements
WATCHES.
FIXES.
LOGS.
Uneven Agent runs 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.
SAFE
CODE
DELIVERY
Write code with absolute peace of mind. Uneven runs continuous local checks as you develop, preventing hardcoded secrets, vulnerability leaks, and supply chain attacks. High-quality programming with built-in security safety nets.
⚠ Code validation is executed locally on your machine without transmitting project data.
Ethical Safety Guard
Proactively blocks the generation of malware, reverse shells, cracking scripts and credential-theft payloads.
Exposed Secrets Prevention
Automatically scans and blocks API keys, tokens, and passwords from being hardcoded in your source files.
OWASP Vulnerability Scanner
Validates that generated code avoids typical security flaws like SQL injections, XSS, and broken rate limiting.
Supply Chain Auditing
Detects suspicious postinstall scripts, typosquatted packages, and obfuscated shell commands in dependencies.
Vulnerable Dependencies CVE Check
Performs instant CVE lookups via public vulnerability databases whenever you modify your project dependencies.
DEVELOPMENT
TO
DEPLOYMENT.
Uneven Agent integrates across your entire software lifecycle, guarding quality, safety, and speed from the first line of code to the production pipeline.
Design & Write
Writing new features, refactoring files, and exploring repository logics.
Queries your codebase context, writes clean code files, and applies surgical refactors via the conversational shell.
Run & Diagnose
Executing local dev servers, compilation steps, and unit test suites.
Runs a background daemon to catch compiler crashes or runtime stack traces in real time, preparing hotfixes instantly.
Validate & Secure
Reviewing code differences and staging files for git commits.
Enforces local guardrails that block hardcoded API keys, malicious package versions, and OWASP code vulnerabilities.
Deploy to Cloud
Running build systems, pull request validations, and deploy pipelines.
Integrates as a headless CI checker (GitHub Actions) to automate malware sweeps and test validations before going live.
FLEXIBLE ENGINES.
LOCAL CONTROL.
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.
Native Engine
RecommendedRuns fully on-device. No API keys, no external network requests, and zero metadata leakage. Optimized natively for secure local repositories.
engine: nativeGoogle Gemini
Cloud EngineSuperior cloud logic with ultra-low latency. Connect your Google AI keys for quick, cost-effective reasoning fallbacks.
engine: cloud apiOllama Daemon
Local EngineIntegrates with any custom model installed in your local Ollama instance (such as Llama 3 or Qwen) for private local tasks.
engine: local daemonOpenAI GPT
Cloud EngineIntegrates GPT-4o models directly into your terminal. Combines cloud intelligence with local project execution boundaries.
engine: cloud apiAnthropic Claude
Cloud EngineConnects Claude Sonnet models. Ideal for deep codebase reasoning and complex multi-file logical refactoring.
engine: cloud apiRegardless of which model is active, Uneven keeps your project indexes, logs, terminal watcher, and file-write boundaries 100% on your machine.
YOUR AI AGENT IS
ONE COMMAND AWAY.
Start free. Local by default. Cloud only if you connect a key — nothing leaves your machine without your say-so.