Installation reference¶

Install AdaMAST for any environment shape: the standard install most readers need, optional provider extras, and a source checkout for contributors. The guided path for new users lives on the documentation home; this page records the dependency variants for repeatable environments.

✅ Requirements¶

Python 3.10 or newer
pip

Note

Installation has no other requirements. Model-provider credentials are needed only later, when you run generation or a model-backed judge; see Providers and models.

📦 Standard installation¶

Install from PyPI. This includes the OpenAI adapter and is the installation path used throughout the quick start:
```
pip install adamast
```

Verify the core commands:

adamast --help
python -m adamast.examples
adamast validate adamast-examples/traces.jsonl

Note

Provider-specific credentials and model selection are documented in Providers and models.

🎛️ Make it yours: provider extras¶

I need models from…	Install
Anthropic	`pip install "adamast[anthropic]"`
Google	`pip install "adamast[google]"`
AWS Bedrock	`pip install "adamast[bedrock]"`

Install only the extra you need, unless the environment needs multiple provider SDKs.

🛠️ Source and development installation¶

For contributors working from a checkout. It is not required to use AdaMAST:

git clone https://github.com/multi-agent-systems-failure-taxonomy/AdaMAST.git
cd AdaMAST
pip install -e ".[test,docs,anthropic,bedrock]"
pytest
python -m mkdocs build --strict

This is the same extras set the release workflow installs; test provides pytest, and docs provides MkDocs Material.

🔌 Host-specific installation¶

General package installation stays here and on the documentation home. The host guides contain only the extra integration steps.

Continue with Codex integration or Claude Code integration.