Installation
Install the Python package, verify the CLI, and prepare Near or SDK use.
Installation
Install the Python package first. Near Desktop embeds the same backend; the public package is still what you use for SDK work and agx on a machine with Python.
What you are installing
Requirements
- Python 3.10+
piporuv- At least one model provider key or a reachable local OpenAI-compatible endpoint (for example Ollama)
Install from PyPI
1pip install agenticx
Optional extras exist in pyproject.toml (desktop runtime, extra LLM adapters, data sources). Only install extras you need. pip install "agenticx[all]" pulls a large optional set and is not required for agx serve plus a single provider.
Install from source
1git clone https://github.com/DemonDamon/AgenticX.git2cd AgenticX3pip install -e .
Verify
1agx --version2agx --help
You should see the agx command group (serve, studio, loop, run, and others). If the command is missing, the console script is not on PATH — reinstall into the same environment you are using.
Provider keys
Put keys in the environment or in ~/.agenticx/config.yaml. One working provider is enough to start.
1export OPENAI_API_KEY="your-key"2# or ANTHROPIC_API_KEY, MINIMAX_API_KEY, and so on
See Configuration for the YAML shape.
Document parsing (optional)
Studio / Near knowledge ingest uses LiteParse for PDF, Office text, and images. It is not the same as the model calling pdftotext in a shell.
1npm i -g @llamaindex/liteparse
.xlsx / .xls need a local LibreOffice. On macOS: brew install --cask libreoffice. If it is missing, the UI should say so and drop those types from the supported list.