### Data Flow Example: Changing max_tokens

Source: https://github.com/itayinbarr/little-coder/blob/main/docs/architecture.md

A step-by-step example illustrating the data flow when a user requests to change the max_tokens configuration.

```text
1. little_coder.py captures input, appends to message list
2. context.build_system_prompt(config) assembles:
   - base + git + memory index + CLAUDE.md
3. local/skill_augment selects tool skills:
   - intent prediction picks Read + Edit
   - injects read.md + edit.md under "## Tool Usage Guidance"
4. local/knowledge_augment scores knowledge files — no match, nothing injected
5. agent.run() streams the model response
6. Model emits: Read({file_path: "config.py"})
   - read_only → auto-approve
   - tool_registry.execute_tool("Read", ...) → file content (truncated if >32K)
7. Tool result appended, next API turn
8. Model emits: Edit({file_path: "config.py", old_string: "max_tokens = 8192",
                      new_string: "max_tokens = 16384"})
   - not read_only → permission request (or auto in accept-all mode)
   - checkpoint.hooks fires, snapshots config.py
   - _edit() runs, generates unified diff
9. little_coder.py renders the diff with ANSI red/green
10. Tool result appended, next turn
11. Model responds: "Done — changed max_tokens from 8192 to 16384"
12. No tool_calls → loop ends, TurnDone yielded
```

--------------------------------

### Local Development Setup

Source: https://github.com/itayinbarr/little-coder/blob/main/README.md

Steps to clone the repository, install dependencies, link the local checkout, and run Little Coder locally.

```bash
git clone https://github.com/itayinbarr/little-coder.git
cd little-coder
npm install
npm link            # makes the local checkout available as `little-coder`
little-coder --model llamacpp/qwen3.6-35b-a3b
```

--------------------------------

### ShellSession Command with Timeout

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/shell_session.md

Example of a ShellSession command with a specified timeout for operations like installations.

```tool
{"name": "ShellSession", "input": {"command": "pip install -q requests", "timeout": 180}}
```

--------------------------------

### BrowserNavigate Tool Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/browser_navigate.md

Example of how to use the BrowserNavigate tool with a specific URL.

```tool
{"name": "BrowserNavigate", "input": {"url": "https://en.wikipedia.org/wiki/Turing_machine"}}
```

--------------------------------

### WebFetch Tool Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/webfetch.md

Example of how to invoke the WebFetch tool with a URL.

```json
{"name": "WebFetch", "input": {"url": "https://docs.python.org/3/library/json.html"}}
```

--------------------------------

### Start Little Coder

Source: https://github.com/itayinbarr/little-coder/blob/main/site/index.html

Command to start the little-coder agent in a project directory.

```bash
lc little-coder v1.5.1 · offline
```

--------------------------------

### Write Tool Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/write.md

Example of how to use the Write tool to create a new Python file.

```tool
{"name": "Write", "input": {"path": "/tmp/example/new_module.py", "content": "def hello():\n    return 'hi'\n"}}
```

--------------------------------

### Install with bun

Source: https://github.com/itayinbarr/little-coder/blob/main/README.md

Install little-coder globally using bun.

```bash
bun add -g little-coder
```

--------------------------------

### Bash Tool Example with Timeout

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/bash.md

Example of executing a shell command with a specified timeout using the Bash tool, suitable for operations like package installations.

```json
{
  "name": "Bash",
  "input": {"command": "pip install requests", "timeout": 120
  }
}
```

--------------------------------

### Serving from another machine on your LAN - Server (Ollama)

Source: https://github.com/itayinbarr/little-coder/blob/main/README.md

Example of starting Ollama server to be accessible on the local network.

```bash
# Ollama: `OLLAMA_HOST=0.0.0.0:11434 ollama serve` (or set `OLLAMA_HOST=0.0.0.0` in the user systemd unit).
```

--------------------------------

### Serving from another machine on your LAN - Server (llama.cpp)

Source: https://github.com/itayinbarr/little-coder/blob/main/README.md

Example of starting llama.cpp server to be accessible on the local network.

```bash
# llama.cpp: start `llama-server` with `--host 0.0.0.0` (or your specific LAN interface) instead of `127.0.0.1`. Everything else from Option A unchanged.
```

--------------------------------

### Install Little Coder

Source: https://github.com/itayinbarr/little-coder/blob/main/site/index.html

Command to install the little-coder agent globally using npm.

```bash
npm install -g little-coder
```

--------------------------------

### Grep Tool Example with Path

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/grep.md

Example of using the Grep tool to find lines matching 'TODO' or 'FIXME' in a specific project path.

```json
{"name": "Grep", "input": {"pattern": "TODO|FIXME", "path": "/path/to/project/"}}
```

--------------------------------

### Bash Tool Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/bash.md

Example of executing a shell command using the Bash tool.

```json
{
  "name": "Bash",
  "input": {"command": "ls -la /path/to/project/"
  }
}
```

--------------------------------

### BrowserType Tool Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/browser_type.md

Example of how to use the BrowserType tool to fill text into an input element and submit.

```tool
{"name": "BrowserType", "input": {"selector": "input[name=\"q\"]", "text": "Alan Turing", "submit": true}}
```

--------------------------------

### Glob Example with Path

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/glob.md

Find all Markdown files in a specific directory.

```tool
{"name": "Glob", "input": {"pattern": "*.md", "path": "/path/to/docs/"}}
```

--------------------------------

### Basic Glob Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/glob.md

Find all Python files recursively.

```tool
{"name": "Glob", "input": {"pattern": "**/*.py"}}
```

--------------------------------

### Single Change Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/edit.md

An example of using the Edit tool to make a single text replacement in a file.

```json
{"name": "Edit", "input": {"path": "/absolute/path/file.py", "edits": [{"oldText": "def hello():\n    return 1", "newText": "def hello():\n    return 2"}]}}
```

--------------------------------

### Grep Tool Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/grep.md

Example of using the Grep tool to find lines matching 'def main' in Python files.

```json
{"name": "Grep", "input": {"pattern": "def main", "glob": "*.py"}}
```

--------------------------------

### Two Changes Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/edit.md

An example of using the Edit tool to make multiple text replacements in a single file within one call.

```json
{"name": "Edit", "input": {"path": "/absolute/path/file.py", "edits": [{"oldText": "MAX = 10", "newText": "MAX = 20"}, {"oldText": "TIMEOUT = 5", "newText": "TIMEOUT = 30"}]}}
```

--------------------------------

### Install with curl

Source: https://github.com/itayinbarr/little-coder/blob/main/README.md

One-line install script for little-coder using curl.

```bash
curl -fsSL https://raw.githubusercontent.com/itayinbarr/little-coder/main/install.sh | bash
```

--------------------------------

### Read Tool with Range

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/read.md

Example of using the Read tool with 'path', 'limit', and 'offset' parameters to read a specific range of lines.

```tool
{"name": "Read", "input": {"path": "/absolute/path/to/file.py", "limit": 50, "offset": 100}}
```

--------------------------------

### BrowserExtract Tool Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/browser_extract.md

Example of how to invoke the BrowserExtract tool with the cursor set to '0'.

```tool
{"name": "BrowserExtract", "input": {"cursor": "0"}}
```

--------------------------------

### Read Tool Basic Usage

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/read.md

Example of using the Read tool with only the required 'path' parameter.

```tool
{"name": "Read", "input": {"path": "/absolute/path/to/file.py"}}
```

--------------------------------

### Bash permission examples

Source: https://github.com/itayinbarr/little-coder/blob/main/README.md

These bash commands show how to configure bash command permissions using environment variables, including adding custom allow-prefixes and skipping the permission gate.

```bash
# Add 'make' (with word-boundary) and 'docker compose ps' on top of the defaults
export LITTLE_CODER_BASH_ALLOW="make ,docker compose ps"
```

```bash
# Skip the gate entirely (use this only inside controlled environments)
export LITTLE_CODER_PERMISSION_MODE=accept-all
```

--------------------------------

### EvidenceAdd Tool Example

Source: https://github.com/itayinbarr/little-coder/blob/main/skills/tools/evidence_add.md

An example of how to use the EvidenceAdd tool to save a source, a note, and a verbatim snippet.

```tool
{"name": "EvidenceAdd", "input": {"source": "https://en.wikipedia.org/wiki/Turing_machine", "note": "Turing machine introduced in 1936", "snippet": "The Turing machine was invented in 1936 by Alan Turing."}}
```

--------------------------------

### Serve Local Model with Llama.cpp

Source: https://github.com/itayinbarr/little-coder/blob/main/site/index.html

Example command to serve a local model using llama.cpp, optimized for GPU usage.

```bash
llama.cpp --n-cpu-moe
```

--------------------------------

### Reproducing the paper (v0.0.2)

Source: https://github.com/itayinbarr/little-coder/blob/main/README.md

Steps to clone the repository, checkout a specific version, and set up the Python environment to reproduce the paper's results.

```bash
git clone https://github.com/itayinbarr/little-coder.git
cd little-coder
git checkout v0.0.2
# Follow that version's README for its Python setup (pip install -e .)
```