A local-first autonomous coding agent powered by Ollama and open-weight models.
Give it a task.
It understands your codebase, plans the work, writes code, runs commands, tests its output, and iterates locally.
Anticlaude is a local autonomous coding agent designed to bring the coding-agent workflow to your own machine.
Instead of sending your source code to a cloud API, Anticlaude uses Ollama and locally running open-weight models to understand your project and work directly inside it.
You give it a goal.
Anticlaude does the rest.
YOUR PROMPT
│
▼
UNDERSTAND THE TASK
│
▼
INSPECT THE CODEBASE
│
▼
CREATE A PLAN
│
▼
WRITE / MODIFY CODE
│
▼
RUN COMMANDS & TESTS
│
▼
DETECT ERRORS
│
▼
FIX AND ITERATE
│
▼
COMPLETE
Why Anticlaude?
Your code stays on your machine
No API calls.
No cloud dependency.
No sending your private source code to a remote model provider.
Anticlaude is built around a local-first workflow.
Bring your own model
Use the models you want.
Anticlaude is designed to work with Ollama-powered local models, allowing you to experiment with different coding models depending on your hardware and requirements.
Qwen
DeepSeek
Code models
Open-weight LLMs
And more through Ollama
No subscription required
Once your local model is running, there is no per-request API billing.
No tokens.
No credits.
No usage meter.
Just your machine and your model.
More than a code generator
Anticlaude is designed around an agentic workflow.
It can reason about a task as a sequence of steps rather than simply generating a block of code.
Analyze
↓
Plan
↓
Execute
↓
Verify
↓
Debug
↓
Repeat
The goal is not simply:
"Generate some code."
The goal is:
Give Anticlaude an objective and let it work toward completing it.
Example
You give Anticlaude a task:
Build a REST API for a task management application.
Use Python and FastAPI.
Add authentication.
Create database models.
Add CRUD endpoints.
Write tests.
Run the tests and fix any failures.
Anticlaude can work through the task by:
[✓] Understanding request
[✓] Inspecting project
[✓] Creating implementation plan
[✓] Creating project structure
[✓] Writing application code
[✓] Creating database models
[✓] Adding API routes
[✓] Writing tests
[✓] Running tests
[✓] Fixing errors
[✓] Task completed
The Philosophy
Modern coding agents are powerful.
But many require:
Cloud access
API keys
Paid subscriptions
Usage limits
Sending code to external infrastructure
Anticlaude explores a different direction.
What if your coding agent lived entirely on your machine?
Your:
computer
models
code
environment
terminal
One local system.
No rented intelligence required.
How It Works
┌──────────────────────────────────────────────────────────┐
│ YOU │
│ │
│ "Add authentication to this application" │
└─────────────────────────────┬────────────────────────────┘
│
▼
┌──────────────────────────────────────────────────────────┐
│ ANTICLAUDE │
│ │
│ Understand → Plan → Execute → Verify → Iterate │
└─────────────────────────────┬────────────────────────────┘
│
┌───────────────┼───────────────┐
▼ ▼ ▼
┌───────────┐ ┌───────────┐ ┌───────────┐
│ FILES │ │ TERMINAL │ │ TESTS │
└───────────┘ └───────────┘ └───────────┘
│
▼
┌──────────────────────────────────────────────────────────┐
│ OLLAMA │
│ │
│ YOUR LOCAL AI MODEL │
└──────────────────────────────────────────────────────────┘
Requirements
Before running Anticlaude, you need:
Python 3.10+
Ollama
A compatible local model
Sufficient system resources for your chosen model
Installation
Clone the repository:
git clone https://github.com/Operandnotfound/Anticlaude.git
Enter the project:
cd Anticlaude
Install dependencies:
pip install -r requirements.txt
Make sure Ollama is running and download a coding model.
Then start Anticlaude:
python runner.py
Example Models
The performance of Anticlaude depends heavily on the local model you choose.
Larger and more capable coding models generally provide better reasoning and code quality but require more powerful hardware.
Experiment with models available through Ollama to find the best balance between:
Speed
Quality
Memory usage
Hardware requirements
Project Goals
Anticlaude is being built toward a future where local coding agents can provide a powerful alternative to cloud-only AI development tools.
The project is focused on improving:
Autonomous task execution
Repository understanding
Multi-step planning
File creation and editing
Terminal execution
Error detection
Self-correction
Test execution
Model flexibility
Local memory
Better CLI experience
Improved agent reliability
Roadmap
Current Focus
Local Ollama integration
Autonomous coding workflow
Project analysis
File editing
Command execution
Task iteration
Future
Multi-agent workflows
Persistent project memory
Better planning systems
Parallel task execution
Web interface
IDE integrations
Plugin architecture
Model benchmarking
Sandboxed execution
Remote/local hybrid mode
Privacy
Anticlaude is built around a simple principle:
Your code should stay yours.
By running the intelligence locally, your source code and development environment remain under your control.
Comparison
Feature Cloud Coding Agents Anticlaude
Runs locally Usually no Yes
Uses Ollama Usually no Yes
Requires API key Often No
Subscription required Often No
Your code leaves your machine Often Not required
Bring your own model Limited Yes
Works with local models Limited Yes
Local-first No Yes
Contributing
Contributions, ideas, bug reports, and feature requests are welcome.
If you find a bug or have an idea that could make Anticlaude better, open an issue.
If you want to contribute code:
git fork
Create a branch:
git checkout -b feature/my-feature
Make your changes and submit a pull request.
Why "Anticlaude"?
Because powerful coding agents should not necessarily require renting intelligence from the cloud.
Anticlaude is not about attacking any particular product.
It's about exploring an alternative:
Powerful coding agents running locally, using models you control, on hardware you own.
<div align="center">
ANTICLAUDE
YOUR CODE. YOUR MACHINE. YOUR MODEL.
Local-first autonomous coding.
<br/>
⭐ If you find Anticlaude interesting, consider giving the project a star.
<br/>
Built for the people who want their AI closer to their machine.
</div>