Open-source framework for AI-assisted software engineering agents
Aget enables developers to create specialized AI agents that collaborate on software projects through discovered patterns rather than rigid commands.
- aget-worker-template - Template for creating worker agents (action-taking or information-only)
- Aget - Framework name (proper noun, capitalized)
- aget- - Prefix for framework components (lowercase)
- -AGET - Suffix for action-taking agents (
⚠️ can modify systems) - -aget - Suffix for information-only agents (✅ read-only)
Examples:
my-github-AGET # Action-taking: Creates PRs, merges code
my-spotify-analyst-aget # Information-only: Reports analytics
The suffix convention provides visual capability signaling - like sudo
or rm -rf
, the capitalized AGET warns of powerful operations.
Privacy-First: Framework works fully offline with no telemetry or required sharing.
Pattern Discovery: Agents learn from interactions rather than following rigid command structures.
Incremental Development: Start with simple capabilities, evolve through use.
-
Clone the worker template:
git clone https://github.com/aget-framework/aget-worker-template my-custom-aget cd my-custom-aget
-
Initialize your agent:
aget init
-
Start working: Wake up your agent and begin collaboration
- Migration Guides: See individual repository docs
- Framework Docs: Coming in v2.4+
- Community: Issues and discussions welcome
Current: v2.4.0 "Clarity"
- Naming conventions established
- Organizational grouping (
aget_group
) - Privacy-first architecture foundation
Framework is in active development. Contribution guidelines coming in v2.5+.
Apache 2.0 (individual repositories may vary - check each repo)
Aget Framework - Discovered patterns, not rigid commands