The AgentExecutor from OnPrem.LLM enables autonomous AI agents to execute complex tasks using both cloud and local models. The pipeline works with any LiteLLM-supported model that supports tool-calling, including cloud models like OpenAI's GPT-5.2-Codex, Anthropic's Claude Sonnet 4.5, and Google's Gemini 1.5 Pro, as well as local models through Ollama, vLLM, or llama.cpp.

Built-in Tools

By default, AgentExecutor provides access to nine built-in tools:

• read_file - Read complete file contents
• read_lines - Read specific line ranges from files
• edit_file - Edit files via find/replace
• write_file - Write complete file contents
• grep - Search for patterns in files
• find - Find files by glob pattern
• run_shell - Execute shell commands
• web_search - Search the web for information
• web_fetch - Fetch and read content from URLs

Configuration Examples

You can customize tool access based on your security requirements:

# Use defaults (all tools including shell):
executor = AgentExecutor(model='anthropic/claude-sonnet-4-5')

Defaults but no shell access (safer):
executor = AgentExecutor(
    model='openai/gpt-5-mini',
    disable_shell=True
)
Minimal tools:
executor = AgentExecutor(
    model='openai/gpt-5-mini',
    enabled_tools=['read_file', 'write_file']
)
Web research only:
executor = AgentExecutor(
    model='openai/gpt-5-mini',
    enabled_tools=['web_search', 'web_fetch']
)

Sandboxed Execution

For security, you can run agents in ephemeral containers using sandbox=True. This is important because agents with shell access can potentially read or modify files outside the working directory. The agent operates within the specified working directory and cannot read or write outside it unless given shell access.

Basic example with sandboxing:

executor = AgentExecutor(
    model='anthropic/claude-sonnet-4-5',
    sandbox=True,
)

result = executor.run(
    task="""
    Create a simple Python calculator module with the following:
    - calculator.py with add, subtract, multiply, divide functions
    - test_calculator.py with pytest tests
    - All tests must pass
    """,
    working_dir='./calculator_project'
)

This approach is useful for developers who need to automate coding tasks while maintaining security boundaries. The tool requires installing PatchPal with pip install patchpal.