Free AI Coding Agent — Claude Code alternative powered by Kimi K2.5. Works on Windows, Linux, macOS.
Built by zero-x Corporation
pip install neuroncli # Windows / macOS / Cloud VMs
pip install neuroncli --break-system-packages # Linux (Ubuntu 23+, Debian 12+, Fedora 38+)# Start interactive REPL
neuron
# One-shot task
neuron "fix the bug in main.py"
# YOLO mode — no permission prompts
neuron --yolo "refactor src/ to use async"
# Plan mode — think first, code later
neuron --plan "add authentication to this API"
# Use local Ollama instead
neuron --provider ollama "explain this codebase"- Type
neuronin any project directory - Browser opens to OpenRouter login (free account)
- API key auto-provisioned + saved to
~/.neuroncli/config.json - Done. No manual config, no copy-paste.
| Feature | Description |
|---|---|
| Parallel Tool Execution | Reads 5 files simultaneously, not one-by-one |
| Git Auto-Commit | Every AI edit is auto-committed like Aider |
| Context Compression | H2A-style compression prevents context bloat |
| 3 Modes | Standard (ask permission), Plan (think first), YOLO (full auto) |
| Session Persistence | Resume previous conversations with /resume |
| Brand Color Output | Clean terminal output with colored tables, headers |
| NEURON.md | Project context file (like Claude's CLAUDE.md) |
| Token Tracking | Shows token usage per session |
| Dual Provider | OpenRouter (cloud, free) + Ollama (local, offline) |
neuron # Standard — asks before writing/running
neuron --plan # Plan — generates plan, waits for approval
neuron --yolo # YOLO — full autonomous, no promptsSwitch at runtime: /mode plan, /mode yolo, /mode standard
| Tool | Description |
|---|---|
read_file |
Read file contents with line numbers |
write_file |
Create or overwrite a file |
edit_file |
Surgical find-and-replace edit |
list_directory |
Tree view of directory contents |
run_command |
Execute shell commands |
search_in_files |
Grep-style search across files |
get_project_structure |
Full project overview |
| Command | Description |
|---|---|
/help |
Show all commands |
/init |
Create NEURON.md project context |
/compact |
Compress context (free up memory) |
/mode <mode> |
Switch: standard, plan, yolo |
/model <name> |
Switch model |
/provider <name> |
Switch: openrouter, ollama |
/upgrade |
Info on faster response times |
/clear |
Clear conversation history |
/config |
Show current configuration |
/exit |
Exit |
| Flag | Description |
|---|---|
--yolo |
Skip all permission prompts |
--plan |
Plan mode — reason before coding |
--model, -m |
Model to use |
--provider, -p |
openrouter or ollama |
--dir, -d |
Working directory |
--no-stream |
Disable token streaming |
neuroncli/
agent.py # ReAct engine — parallel tools, context compression
auth.py # OpenRouter OAuth auto-provisioning
cli.py # REPL + slash commands
config.py # Modes (standard/plan/yolo)
git_integration.py # Auto-commit AI edits
session.py # Save/resume conversations
prompts.py # Mode-aware system prompt builder
tools.py # Tool registry + 7 tool implementations
ui.py # Brand colors, startup screen
provider.py # Provider abstraction layer
openrouter_client.py # OpenRouter (Kimi K2.5)
ollama_client.py # Local Ollama
tests/
test_agent.py # 18 unit tests
- Python 3.10+
- Internet connection (for Kimi K2.5 via OpenRouter)
- Optional: Ollama for local offline mode
MIT — free to use, modify, and distribute commercially.
By zero-x Corporation — Tamil Nadu, India