Astra Agent is now open source. The full control plane, the TUI, the ARA SDK, and the deployment tooling — all of it lives at github.com/jcplaboratory/astra-agent.
Why now
Astra has been running as my personal agent for months — handling email, monitoring infrastructure, managing tasks. It works. The architecture is stable. The memory pipeline is governed and auditable. The ARA delegation model has proven itself.
Open-sourcing now means the community can inspect the security model, contribute specialist ARAs, and run their own instances. Privacy-first AI shouldn't be a black box.
What's in the repo
- Control Plane — FastAPI, MariaDB, Qdrant, S3/MinIO, OIDC + mTLS
- Memory Pipeline — structured extraction, candidate review, provenance tracking
- ARA SDK — bounded remote agents with capability-scoped leases
- Textual TUI — terminal interface with conversation, task, and memory views
- Interactive Installer —
bash install.shfor a full local stack
What it's not
Astra is not a chatbot platform, a plugin marketplace, or a "deploy anywhere" SaaS. It's a personal agent that runs on infrastructure you control. There's no telemetry, no vendor lock-in, and no API key that expires when funding dries up.
Getting started
git clone https://github.com/jcplaboratory/astra-agent.git
cd astra-agent
uv sync --all-packages
uv run astra-agent
The README has the full setup guide. If you run into issues, open a GitHub issue — I read them all.