13 projects
bastion-broker
A single-GPU safety broker for Ollama — serializes model loads and budgets VRAM to prevent crashes on a shared GPU. Not a serving engine.
tallyhouse
Reads Claude Code session transcripts into typed pointers and verbatim spans — transcript access and parse-coverage.
guildhall
Multi-agent orchestrator and memory substrate — the parent system the daughter packages are cut from.
session-ledger
Session recaps and digests for coding-agent work — consolidates past sessions into what to pick up next.
dispatch-rider
Context retrieval and injection for AI-agent runs.
stenopipe
Local voice dictation — speech transcribed on your GPU and piped wherever you point it.
muster-roll
Context, approval, disclosure and settle layer for AI-agent dispatch — what was seen, approved by whom, and what fired.
countinghouse
Reads Claude Code session transcripts into typed pointers and verbatim spans — transcript access and parse-coverage.
evidencebook
Cross-session content recall over Claude Code session transcripts.
marshalyard
A trust layer of skills and workflows for unattended, multi-agent coding-agent runs.
knowledge-keep
A tiered episodic memory store for AI agents.
dag-dispatcher
A DAG router for multi-agent task orchestration.
weighhouse
Admission jury and memory-review / oversight UI — the write-time gate of a multi-agent memory system.
