The supplied case attributes Engrim to creator timgordontg, whose Show HN presentation calls it a “universal, local-first SQLite memory engine for AI CLIs.” Its proposed significance is shared persistent memory across CLI tools rather than separate tool-specific stores. None of the supplied web results directly identifies Engrim or its creator: they describe OpenMind, OpenClaw, similarly named Engram projects, and other memory tools, so they do not independently establish Engrim’s release status, implementation, or cross-CLI compatibility.
Engrim’s claimed cross-CLI persistence repeats Scott’s surface-independent memory position in “Retail MCP Is the Doorway Not the Memory”; the radar already follows similar claims in MemHub and OpenContext, though neither page establishes prior tracking of Engrim itself. With only the creator’s presentation and no independently established compatibility or implementation details, this adds another example—not evidence that would change Scott’s architecture or establish a consequential adoption of his position.
ip:source.retail-mcp-is-the-doorway-not-the-memory-ebookradar:memhub-shared-coding-agent-memoryradar:opencontext-project-local-agent-memory
queries asked of Scott's wikis
- shared agent memory across coding harnesses and CLIs
- persistent project knowledge versus tool-specific memory
- local-first SQLite agent memory architecture
- agent-maintained wikis versus retrieval memory stores
- portable agent context and harness switching
2026-09-10T23:40:50Z
The announcement has exhausted its monitoring window without new implementation evidence or a concrete forthcoming validation milestone. Expire the episode as faded, not disproved; a reproducible cross-CLI memory demonstration would justify reopening it.
2026-09-08T22:43:08Z
The refreshed discussion adds general skepticism about memory quality, not an Engrim-specific failure report or successful implementation evidence. The creator’s architectural rationale does not settle cross-CLI interoperability, write triggers, or retrieval usefulness, so the case remains an unvalidated portable-memory implementation.
2026-09-07T15:25:03Z
The refreshed comments repeat compatibility and memory-write questions without supplying answers or independent usage evidence. Engrim remains a creator-announced portable-memory implementation, not yet evidence of reliable shared persistence across coding harnesses.
2026-09-07T14:37:22Z
The refreshed evidence adds no substantive change beyond previously reviewed setup claims and unanswered compatibility questions. Engrim remains a creator-announced implementation, with neither independent cross-CLI validation nor evidence that would change Scott’s memory architecture decisions.
2026-09-07T13:23:26Z
A commenter quotes setup instructions claiming automatic detection and configuration of installed environments, adding a concrete integration claim but no successful cross-CLI demonstration. The accompanying uninstall question makes setup reversibility another evaluation concern, not an established defect; memory-write behavior and interoperability remain unvalidated.
2026-09-07T09:29:45Z
The refreshed discussion adds OpenCode to the unanswered compatibility requests, not evidence that Engrim delivers shared memory across harnesses. The case remains a creator-announced implementation with unvalidated interoperability and memory-write behavior.
2026-09-07T08:27:48Z
The refreshed discussion identifies the practical tests Engrim still needs to pass—Codex/Pi compatibility and how memories get written—but supplies no answers or usage validation. Interest from an existing provider-agnostic memory user signals demand, not adoption or independent corroboration of Engrim.
2026-09-07T05:26:50Z
No substantive delta changes the interpretation: Engrim remains a creator-announced portable-memory tool, not demonstrated cross-CLI interoperability. The GitHub echo repeats the same announcement rather than providing an independent line of evidence.
2026-09-07T05:25:21Z
grounded: known/low — Engrim’s claimed cross-CLI persistence repeats Scott’s surface-independent memory position in “Retail MCP Is the Doorway Not the Memory”; the radar already foll
2026-09-07T05:22:46Z
case created — A concrete author-linked implementation warrants a bounded capability case, though interoperability and operational reliability remain unsubstantiated.