OPEN SOURCE · MCP NATIVE · YOUR FILES
Keep the findings, decisions, and hard-won context your AI would otherwise forget. HyperKB turns them into connected, searchable memory you own.
The reason stays with the decision.
Record what you learned, together with why it matters.
Good context shouldn't depend on one conversation.
FROM A MOMENT TO A MEMORY
HyperKB gives an agent a place to put what matters—and a way to find it again when the work calls for it.
A finding, a decision and its rationale, or a task for later. Add a topic, a status, and links to related knowledge.
Dotted filenames organize topics. Wiki-links connect related ideas. The Markdown is yours to read, back up, and move.
Search with ripgrep and BM25 ranking, build a session briefing, or pack relevant entries into an estimated context budget.
The decision and its original rationale, together.
USEFUL MEMORY. ORDINARY FILES.
Keep a knowledge base outside your conversations. Connect your MCP clients to it, carry it between machines, and read it without an AI tool at all.
Understand the storage modelYour knowledge lives in named text files. SQLite is an index you can rebuild.
Local MCP clients share the knowledge base while keeping their focus and provenance separate.
Keep recording locally through outages and upgrades. Optional S3 sync checks compatibility before sharing your pending changes.
MAKE THE NEXT SESSION BETTER
Install HyperKB, connect your MCP client, and save a decision worth remembering. No model API key is needed for local search.
git clone https://github.com/calvincs/hyperkb
cd hyperkb
python3 -m venv .venv
.venv/bin/pip install -e ".[mcp]"
.venv/bin/hkb inithkb-mcp in your client.A SHORTER PATH TO UNDERSTANDING