Project summary

Over the next six months, Mark Hopkins will advance SanctumOS (sanctumos.org), a self-hosted cognitive operating system that serves as open infrastructure for agentic AI—similar to how Linux established the foundation for modern operating systems. By enabling reproducible research automation, scientific assistants, and interdisciplinary collaboration, SanctumOS directly contributes to the advancement of science and technology.

SanctumOS is organized around a neuro-inspired architecture, where core subsystems echo cognitive functions. The Thalamus refines raw input streams, the Cerebellum filters and prioritizes signals, and Broca translates agent decisions into communication. The Dream Agent is further structured into Hippocampus, Neocortex, and Glymphatic modules—distilling daily activity into knowledge, integrating it into long-term structures, and compacting indexes for efficient retrieval. This framing makes the system both intuitive and rigorously modular.

At the same time, SanctumOS tackles technical AI safety through design: every agent runs under user control, with persistent memory, transparent provenance, and open-source governance. This reduces systemic risks from centralized, black-box AI by ensuring agents remain auditable, accountable, and self-hostable.

What are this project's goals? How will you achieve them?

Goal 1: Establish a Neuro-Inspired Multi-Agent Architecture

How we’ll achieve it: We will conduct iterative research and prototyping focused on developing and validating the core SanctumOS multi-agent framework — including the Hippocampus, Neocortex, and Glymphatic modules of the Dream Agent, the Thalamus and Cerebellum routing/processing layers, and the Broca 2 communication middleware. This will involve technical validation of memory management strategies, agent-to-agent orchestration, and simulated chain-of-thought reasoning for small local models, ensuring the outcome is both functional (robust cognition, verifiable provenance) and sustainable (efficient, CPU-friendly, open-source ready).

Goal 2: Build Supporting Tools and Integration Infrastructure

How we’ll achieve it: By building complementary sub-agents, tooling APIs, and developer documentation, we will support integration and usability. This includes:

• Tool-layer services (search, citation management, summarization, indexing, markdown linting) for the Deep Research Division.

• Refining MCP governance and contracts between modules (Hippocampus ↔ Neocortex ↔ Glymphatic, Supervisor ↔ Researcher, Companion ↔ Deep Research).

• Enhancing interoperability with external ecosystems (LangChain, LlamaIndex, local model runtimes like llama.cpp).

• Producing developer guides, naming conventions, and structured rubrics so the system is teachable and extensible for future contributors.

• All code is published under GPLv3 and all supporting documentation is published under Creative Commons (CC-BY-SA).

Goal 3: Deliver a Prototype Demonstration and Open Release

The project will culminate in a prototype deployment of SanctumOS with working Dream Agent nightly consolidation and Broca 2 multi-channel communication. Crucially, this will not be a single installation, but a dummy-proof install process that allows anyone to deploy SanctumOS—locally on their own machine, across a home network, or on a VPS. By making end-to-end private, personal AI accessible, we enable researchers, practitioners, and individuals to run their own agents with full control. A community-facing open-source launch, demo event, and pilot integrations (e.g., summarizing long conversation logs, generating structured dossiers, producing citation-rich reports) will validate SanctumOS as tangible, reproducible, and accessible infrastructure for knowledge work. In short, we want to create an e2e solution that is a competitive product with quality on par with industry leaders like ChatGPT or Claude.

How will this funding be used?

Our goal of $16,000 would give us two months of runway, putting us well into a solid 1.0 release territory while also using this funding to accelerate core development for the Sanctum platform.

Our minimum funding amount will keep our lights on while we continue to put out useful agentic AI open source repositories (http://github.com/sanctumos)

Resources will support the building, testing, and refinement of modules such as Broca, Thalamus, and the Engagement Protocol. This includes covering engineering time, infrastructure for testing and deployment, and integration work to ensure the system is robust, extensible, and ready for release.

In parallel, a portion of the funding will be dedicated to developer relations. This will involve producing documentation, onboarding guides, and open-source support systems to make Sanctum accessible to external developers. We will also run workshops, prepare example integrations, and maintain community-facing channels, ensuring that contributors can effectively engage with the project and help grow the ecosystem.

A portion of funds will also be allocated toward nonprofit incorporation and compliance costs, ensuring Sanctum can operate with the appropriate legal structure to steward its open-source mission and manage future grants effectively.

Who is on your team? What's your track record on similar projects?

Sanctum is led by Mark Hopkins, a veteran technologist with two decades of experience in decentralized systems, peer-to-peer infrastructure, and open-source collaboration. Mark has previously served as Operations Director of FreeRossDAO (informing the governance design of AssangeDAO and UkraineDAO), contributed to ThresholdDAO and other Web3 governance systems, and has shipped production software across AI, cryptography, and distributed media.

While Hopkins is the primary developer, Sanctum has already attracted a community of 25–50 contributors who have provided code segments, branding, web design, and user feedback. This distributed, open-source approach is intentional: SanctumOS is designed as infrastructure anyone can install and extend, and the contributor base ensures early validation and continuous iteration.

There's also Jess Hopkins who helps with writing these proposals.

What are the most likely causes and outcomes if this project fails?

If this project does not succeed, the most immediate outcome is that while the open-source codebase will remain available to the public, it will not be developed into a fully realized, end-to-end privacy-focused AI system. The technical contributions would still exist as reusable building blocks, but they would not reach the stage of becoming a cohesive platform that demonstrates a smaller, independent alternative to today’s dominant AI infrastructures.

The broader consequence is that without Sanctum, the AI ecosystem will continue to consolidate around large, centralized players. These companies already control most of the tooling, models, and deployment pipelines. Failure to establish Sanctum as a working alternative would mean that independent developers, researchers, and communities lose an opportunity for greater autonomy and privacy in how AI is built and used. In other words, the “default future” tilts more toward centralized control, whereas this project represents a chance to diversify the ecosystem with a user-first, open, and privacy-preserving approach.

How much money have you raised in the last 12 months, and from where?

SanctumOS has not yet taken outside investment. Instead, Mark Hopkins has personally subsidized development with over $20,000 of his own funds over the past year. This has covered both his full-time commitment and contributions from community collaborators (code, design, and feedback). That bootstrap investment has enabled Sanctum to progress from concept to a working multi-agent framework with an active contributor base—demonstrating traction before seeking external support.