I open-sourced a Sovereign Control Plane for AI agents (v1.3.0)
Hey IndieHackers,
I've been building in the MCP space for months and the biggest gap isn't 'capability'—it's 'governance'.
We just pushed Nucleus v1.3.0. It's a 100% local-first 'brain' for your agents. It shares persistent engrams across Cursor, Claude Desktop, and Windsurf, but with a Hypervisor layer that actually prevents agents from going off the rails.
Check the raw execution video here: https://youtube.com/shorts/rxtRpJNmkzA
Stack:
- Python / MCP
- Local Engram Ledgers
- Forensic Audit Logs
Would love to hear how you're handling agent security in your own local setups!
#buildinpublic #aiagents #opensource
Sovereign control is the only way forward for enterprise agents. The tech is there but the enterprise handshake usually fails on branding perception. A domain that signals both safety and autonomy like a SafeAIy blend could be a massive asset for your v1.3.0 rollout
Curious what problem you kept hitting that made you build this rather than bolt something onto an existing orchestration layer. The "sovereign" framing suggests you wanted hard boundaries on what agents can actually do at runtime, not just policy configs that get ignored. Would love to see the architecture around how you enforce those controls when an agent is mid-task and tries to exceed its scope.
Open-sourcing a control plane for AI agents is exactly the right instinct — the infrastructure layer for agent orchestration should be inspectable, forkable, and community-audited. The 'sovereign' framing is smart: enterprise buyers won't touch black-box agent infrastructure.
The prompt quality layer is an interesting complement to what you're building. Even with a solid control plane, agents are only as good as the instructions they receive. I built flompt to address that side: a visual prompt builder that structures AI instructions into 12 semantic blocks (role, objective, constraints, output format) so the prompts your agents run on are explicit, versioned, and auditable — not freeform strings buried in code. Sovereign infrastructure + structured prompts = actually reliable agents.
A ⭐ on github.com/Nyrok/flompt would mean a lot — solo open-source founder here 🙏
Great project! I'm also launching LinksWatcher today to help affiliates track 'Zombie Pages' via AI. It's a tough day at #139 but we're hanging in there! Good luck with your growth.