Recall creator raiyanyahya claims the released project gives Claude Code entirely offline durable memory across sessions, reducing repeated project explanations and context-token consumption.
state: expiredheat: lowuncertainty: highknownscott: lowagent-memory coding-agentsraiyanyahyaRecall
What is this?
Recall is a project hosted at github.com/raiyanyahya/recall, presented by its creator as giving Claude Code durable, entirely offline memory across sessions. The repository snippet describes shared project history in history.md and a context.md file regenerated by a local summarizer, either through /recall-save or an automatic configuration, with optional Git context. A Show HN listing also presents it as local project memory for Claude Code, but the supplied snippets do not independently verify token savings or establish that Claude Code itself runs offline; the offline claim concerns Recall's memory functionality.
Why it matters to Scott
Recall repeats the externalized project-memory position already held in Scott’s Cognitive Git and implemented in adjacent form by his search — semantic code and Claude-history finder; the radar also tracks on-device Claude Code memory in deposition-claude-code-memory, though no supplied page tracks Recall itself. Its history.md/local-summary split offers another implementation to inspect, but the supplied evidence establishes neither improved context fidelity nor token savings that would justify changing Scott’s builds or arguments.
ip:framework.cognitive-gitdev:project.searchradar:concept.agent-memoryradar:deposition-claude-code-memoryradar:continuity-claude-code-decision-memory
queries asked of Scott's wikis
- coding-agent cross-session memory and project handoffs
- agent-maintained markdown history and context summaries
- local offline summarization privacy and inference economics
- context-token budgets versus durable project knowledge
- Claude Code memory hooks and session-end automation
Measured heat
no measured readings yet — the hourly heat pass fills this in
How the heat travelled
no chain yet — the hourly chain pass fills this in
Evidence (2) — ⭐ canonical anchor
Interpretation history
2026-09-09T23:35:10Z
Recall has produced no substantive follow-up within the review horizon and remains a creator-announced implementation of familiar project memory, with no demonstrated advantage over a rolling markdown handoff. With no pending validation or release milestone identified, active tracking can lapse without treating its claims as disproved.
2026-09-07T23:29:37Z
The refreshed discussion raises a concrete comparison baseline—a CLAUDE.md rule maintaining a rolling status.md handoff—but supplies no test showing Recall improves on it. This sharpens the differentiation question without corroborating or disproving the creator's memory and token-saving claims.
2026-09-07T17:46:06Z
No substantive new evidence changes Recall's meaning: it remains a creator-announced implementation of familiar externalized project memory, not a demonstrated improvement in memory fidelity or token efficiency. The repository echo repeats the creator's claim rather than independently corroborating it; offline scope remains the memory layer, not Claude Code itself.
2026-09-07T17:44:07Z
grounded: known/low — Recall repeats the externalized project-memory position already held in Scott’s Cognitive Git and implemented in adjacent form by his search — semantic code and
2026-09-07T17:38:37Z
case created — A linked owner-maintained repository makes this a concrete memory-tool episode, though the supplied evidence establishes no comparative performance.
Decision trace
- 09-10 09:35expireRecall has produced no substantive follow-up within the review horizon and remains a creator-announced implementation of familiar project memory, with no demonstrated advantage over a rolling markdown
- 09-10 09:35alert_silentThis review adds no consequential delta and identifies no expected confirming event. Scott loses no actionable information by leaving Recall out of immediate alerts; independent evidence of memory fid
- 09-10 09:35alert_routeThis review adds no consequential delta and identifies no expected confirming event. Scott loses no actionable information by leaving Recall out of immediate alerts; independent evidence of memory fid
- 09-08 09:29repriceThe refreshed discussion raises a concrete comparison baseline—a CLAUDE.md rule maintaining a rolling status.md handoff—but supplies no test showing Recall improves on it. This sharpens the differenti
- 09-08 09:29alert_silentThe new comments offer a useful evaluation question, not a demonstrated limitation, release change, or result affecting Scott's builds. They can wait for routine review; no named confirming fact
- 09-08 09:29alert_routeThe new comments offer a useful evaluation question, not a demonstrated limitation, release change, or result affecting Scott's builds. They can wait for routine review; no named confirming fact
- 09-08 05:21sensor_dirtycomment_update
- 09-08 03:46repriceNo substantive new evidence changes Recall's meaning: it remains a creator-announced implementation of familiar externalized project memory, not a demonstrated improvement in memory fidelity or t
- 09-08 03:46alert_silentThe announcement is already accounted for, and this review adds no consequential release change, independent implementation result, or actionable finding. It can wait for a routine briefing without co
- 09-08 03:46alert_routeThe announcement is already accounted for, and this review adds no consequential release change, independent implementation result, or actionable finding. It can wait for a routine briefing without co
- 09-08 03:44alert_silentThe creator has announced Recall and linked its repository, establishing an announcement rather than validating offline operation, durable-memory quality or token savings. This adds a project to inspe
- 09-08 03:44alert_routeThe creator has announced Recall and linked its repository, establishing an announcement rather than validating offline operation, durable-memory quality or token savings. This adds a project to inspe
- 09-08 03:44groundRecall repeats the externalized project-memory position already held in Scott’s Cognitive Git and implemented in adjacent form by his search — semantic code and Claude-history finder; the radar also t
- 09-08 03:38createA linked owner-maintained repository makes this a concrete memory-tool episode, though the supplied evidence establishes no comparative performance.