Local-first memory, skills and tracing for your agent: engram (recall/remember your notes via MCP), skillet (search/install agent skills via MCP), plus skills that teach the agent to use them — nothing leaves your machine.
Four-layer long-term memory (L0 Conversation → L1 Atom → L2 Scene → L3 Persona) for Claude Code — local FTS5 + EmbeddingGemma vector hybrid recall, agent-driven extraction and consolidation, zero external API.
Interview prep skills: role intel, a sourced story bank, speakable verbatim scripts, industry briefs, mock interviews with scoring, and debriefs that feed the next round.
Persistent memory for AI assistants — runs locally on your machine. Remember every decision, debugging session, and architecture debate across sessions. One MCP server, zero cloud dependencies, works on a Raspberry Pi.
Production debugging bundles for AI agents. Connect Claude Code to DebugBundle incidents, deterministic bundles, reproductions, health checks, probes, alerts, webhooks, projects, and verification workflows.
A disciplined control loop for multi-window Claude Code work: a controller window dispatches scoped task packages across a tmux-resident fleet, collects immutable target results, obtains independent Test validation, reviews the evidence, and decides the next step. Explicit on-disk state roots keep the run auditable…
Typed, local-first memory for Claude Code — decisions, domain rules, and runbooks Claude reads on its own. Human-confirmed writes, read-only MCP, benchmarked supersession-aware recall. Branch/PR-aware and team-shareable.
Offload summarisation, triage, classification, and bulk-text tasks from Claude Code to locally-installed models. Saves Anthropic API tokens and keeps content on-device.