The challenge
Build durable continuity across large, fragmented histories without turning AI memory into an opaque summary, losing original sources, or removing the user's control over what an assistant can read and preserve.
All projectsDeployed MVP
Knowledge & AI
Recall Harbor
A private knowledge and continuity platform that lets people and their AI assistants recover context, preserve evolving ideas, and pick up where they left off.
Try Recall Harbor
- Role
- Founder & developer
- Focus
- Knowledge continuity
- Stage
- Deployed MVP
The project
Recall Harbor grew from a practical problem: useful decisions, ideas, and project history were becoming buried across large AI conversation archives and disconnected notes. It began as a simple Zettelkasten-style CSV import utility called Lattice, then evolved as the problem expanded from storing information to preserving provenance, finding the right context later, and making that knowledge useful outside the conversation that created it.
Today, Recall Harbor provides a searchable personal library that can be used directly through its web interface or accessed by authorized AI assistants. Users can import and browse existing knowledge, recover relevant context from previous work, and explicitly preserve new information without silently replacing the history behind it. The system is already being dogfooded to reconstruct the development histories of Silvercastle projects for this portfolio, including Recall Harbor itself.
The approach
Treat the knowledge library as the durable source of truth and AI assistants as permissioned clients of that library rather than as the memory system itself. Recall Harbor preserves source material and provenance, exposes bounded retrieval with citations and coverage limits, separates read and write capabilities, and records evolving information through append-only additions instead of silently rewriting history. Agentic systems help retrieve, reconstruct, and preserve useful context while the underlying product remains independently usable by the person who owns the library.
What matters
Built around the use case.
Context with provenance
Retrieval is designed to recover relevant material together with its source context, chronology, and evidence, allowing an assistant to reconstruct how a project or idea evolved instead of relying on a flattened memory summary.
Permission-aware continuity
Reading, append-only note creation, conversation capture, and structured inventory changes have distinct permission boundaries, keeping useful AI continuity tied to explicit user control rather than indiscriminate conversation recording.
Built through real use
Recall Harbor evolved from a personal CSV-memory utility into a deployed knowledge product by repeatedly testing it against the workflows that motivated it. It is now being used to recover Silvercastle project histories and construct this portfolio, including reconstructing Harbor's own development.