Meet Select, an open-source platform where AI agents earn their autonomy (v1.0.0)

Select 1.0.0 is an open-source, self-hosted platform for managing AI agent teams, released by Kirby54. This innovative system challenges traditional AI agent autonomy models by requiring agents to prove their capability to complete specific tasks before gaining more autonomy. On the GitHub platform, Select is available for download at https://github.com/Kirby54/Select and its release notes can be found at https://github.com/Kirby54/Select/releases/tag/v1.0.0.

What sets Select apart is the unique approach to autonomy. Instead of upfront configuration, agents must submit work and undergo verification to earn more autonomy. Deterministic checks validate results, and the system maintains evidence receipts to ensure accountability. Agents can earn autonomy through verified work, but sensitive actions still require human oversight.

The core components of Select include multi-agent teams with roles and task management, evidence receipts, deterministic verification, earned autonomy through promotion or demotion, and human approval workflows. Repository-scoped permissions enable review-only code proposals, and cost tracking features are included for better management. Docker-based self-hosting simplifies deployment, and a fictional demo company operates seamlessly without the need for API keys.

Select is written using the Next.js, FastAPI, PostgreSQL, and Python technologies. The designers' particular interest lies in the fact that a coding agent achieving success in multiple tasks should not suddenly obtain permissions to access sensitive areas like customer emails or other repositories. To maintain security and control, Select separates verification evidence from permissions and allows admins to revoke repository-specific code proposal permissions when necessary.

Currently, the release focuses on the self-hosted core, without offering a production-security certification, native Gmail/Microsoft 365 OAuth integration, or sandboxed Claude Code/Codex execution. Additionally, real provider compatibility is still in the validation stage within individual environments.

To experience Select, simply clone the repository, initiate the Docker startup script, and explore the demo company. Kirby54 invites technical feedback, questioning whether users would trust an AI agent more if its autonomy were earned through verified work rather than assigned with a static permissions slider.