Zero-dependency context survival kit for Claude Code. Auto-checkpoints your session state to plain Markdown before compaction, restores it after — plus focus skills that keep Claude lean and on-task. No database, no background process, no AI compression. Just files you can read.
★not rated 2 2mo agoA
tokens not measured
originalMIT
Skills for working with Heptabase via the CLI. Bundles a general CLI skill (notes, journals, tags, cards, whiteboards, AI Tutor) and a focused cross-linking skill that explains the ProseMirror card/date node schema and ships a helper script.
Autonomous state machine for cognitive continuity. Preserves session state before context compaction and restores it on session start. Features priority-based selective restore, token budget tracking, auto decision extraction, and state archiving. Based on Alan Turing's 1936 paper 'On Computable Numbers'.
★not rated 2 7mo agoA
tokens not measured
originalMIT
Durable project state across AI sessions: decisions, constraints, and rejected paths, versioned in git as plain markdown. Not a memory tool — it answers what is true about the project now, and what must not change.
★not rated 2
changed yesterdayA
tokens not measured
originalMIT
Memory for Claude Code that reasons about you and improves itself: Honcho-pattern belief store with deriver, dreamer and dialectic (/lore:ask), automatic skillification with a closed judge/update/retire loop, on a Hermes-pattern core — hard-capped curated memory, FTS5 session search, staged review.
Most explorations die with the session. FragCap captures them as structured capsules — the problem, the dead ends, the open threads — and makes them searchable by anyone starting something similar. Your unfinished thinking becomes their shortcut.
★not rated 2 5mo agoA
tokens not measured
originalMIT
A memory that grows. Your agent is brilliant within a session and amnesic between two — this gives it a persistent knowledge trunk on your machine: what you work out once is distilled into a note, filed, linked, and surfaced automatically next time, from any project. It makes no request of its own — no telemetry, no…
★not rated 2 11d agoA
tokens not measured
originalApache-2.0
Bifrost MCP gateway plugin for Claude Code and Claude Desktop: skill discovery, agent-driven memory, one-command onboarding, signed admin policy, and OAuth 2.1 for Desktop.
★not rated 2 18d agoA
tokens not measured
originalMIT
A cheap background model that watches your Claude Code session over its shoulder and, rarely, drops one short advisory into it — asynchronously, without ever blocking. It reads only the agent's emissions (29% of the transcript), never its tool results, so it is affordable to leave running for the whole session.
★not rated 2 9d agoA
tokens not measured
originalApache-2.0
Warns the main session as its own context fills, captures a wake packet before compaction, and reorients the session after it — so an autonomous orchestrator survives past one context window.
★not rated 2 9d agoA
tokens not measured
originalMIT
Local-first codebase memory for coding agents: AST-aware context, durable decisions, failure records, review packs, and a local graph GUI. Twenty tools are advertised by default; writefile, patchfile, prompt compression, and bash are opt-in. Discipline protocols ship as MCP prompts rather than tools. Pre-2.0 tool nam.
★not rated 2 3d agoA
tokens not measured
originalMIT
Closes the memory loop mechanically: recall before changing code, recall before storing, and a turn that cannot end with an unmade decision left silent. Requires Node >= 23.6 (the hooks run TypeScript directly).
★not rated 2 1mo agoA
tokens not measured
originalApache-2.0
Cross-session, cross-device memory for Claude Code. Markdown-first captures, vector-backed recall. Points at any universal-memory server — local Docker, remote self-hosted, or cloud-backed. Auto-starts the server on demand when configured to.
★not rated 2
changed yesterdayA
tokens not measured
originalMIT