Skip to content

Quickstart

Install

Python 3.10+ is required.

python -m pip install "dashscope-agenticx[tui]"
agenticx --version

For a minimal, non-TUI installation:

python -m pip install dashscope-agenticx
agenticx --cli
python -m venv .venv
# macOS / Linux
source .venv/bin/activate
# Windows PowerShell: .venv\Scripts\Activate.ps1
python -m pip install -e ".[tui]"
agenticx --version

The distribution is dashscope-agenticx, the import namespace is dashscope_agenticx, and the only official command is agenticx. The bare agenticx PyPI distribution is a different project.

Optional extras

Extra Purpose
tui Textual interface with fixed input panel and rich output
provider-openai, provider-anthropic Provider-specific SDK routes
web, browser Search and browser tools (browser runtime setup is separate)
voice, camera, vision (or media) Audio, camera, and image dependencies
dev, all Test/build tooling and all optional features
python -m pip install "dashscope-agenticx[web,browser]"

Installing dependencies does not grant an App permissions.

Set an API key

For the default Tongyi/DashScope provider:

export DASHSCOPE_API_KEY="<your-api-key>"
# Windows PowerShell: $env:DASHSCOPE_API_KEY="<your-api-key>"

Other providers use ANTHROPIC_API_KEY / OPENAI_API_KEY, or the per-provider key slots managed by /provider (stored in the OS keyring). See Configuration. Do not paste real keys into tracked configuration files.

First terminal run

agenticx

A new workspace without .agenticx/config.toml starts the /setup wizard: username, configuration mode (Default — Alibaba/Bailian, or Custom), provider/model, API key, capabilities, and an optional expert template. Existing configured workspaces skip straight to the prompt.

agenticx> Summarize this project's structure before making any edits
agenticx> /help
agenticx> /provider
agenticx> /session

Other entry modes:

agenticx --cli                      # inline REPL with native scrollback
agenticx --tui                      # Textual interface (tui extra)
agenticx -c "List the files here and summarize the project"
agenticx --dry-run                  # preview loaded configuration

One-shot runs can execute tools. Start in a suitable workspace and keep the default confirm_mode = "all" until you trust the task.

First SDK call

from dashscope_agenticx.sdk import run_once_sync

reply = run_once_sync("Summarize the key points of README.md")
print(reply)

run_once_sync runs a complete turn — including tool calls — and returns the final assistant text. For streaming and typed events see Embedding the SDK.

Next steps