2026-10-11 17:11 UTC

Independent use will determine whether Anarlog provides a practical privacy-preserving meeting-memory workflow through on-device transcription, pluggable summary models, local retrieval, and MCP access.

state: expiredheat: lowuncertainty: highknownscott: mediumlocal-inference agent-memory rag-knowledgeAnarlog

What is this?

Anarlog is an MIT-licensed, open-source meeting notetaker positioned as an alternative to Granola: it captures audio without joining calls, performs transcription on-device, and stores sessions, transcripts, notes, recordings, and attachments locally, including in SQLite and plain files. Its site says users can run local models, bring their own API keys, or use Anarlog’s AI, while Markdown export supports portability; the project was previously named Char, and the evidence identifies yujonglee as author of the repository’s initial December 2024 commit. The supplied snippets do not independently establish the quality of retrieval, the practicality of the complete workflow, or the nature and maturity of MCP access, so those remain claims requiring hands-on testing.

Why it matters to Scott

The core architecture is already held across Model Perishability and Scott’s active Ambient Conversation Copilot, local speech, model-routing, and MCP-wiki work. Anarlog adds a potentially reusable integrated implementation rather than a new position; hands-on validation could still affect what Scott builds by testing whether local transcription, portable meeting records, swappable summarizers, retrieval, and MCP access work coherently in one privacy-preserving workflow.
ip:concept.model-perishabilitydev:project.listendev:concept.hardware-aware-local-inferencedev:concept.task-aware-model-routingdev:project.mcp-ip-wikidev:technology.whisperradar:concept.agent-memoryradar:concept.local-inferenceradar:concept.privacyradar:concept.model-routingradar:concept.mcpradar:concept.knowledge-systems
queries asked of Scott's wikis
  • local-first agent memory architecture
  • private meeting memory and personal knowledge capture
  • on-device transcription and local inference economics
  • pluggable model providers and bring-your-own-model design
  • SQLite retrieval for agent-maintained knowledge
  • MCP access to personal memory stores

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

sourceobjectauthorscorecomments
🟠 redditOpen-source meeting notetaker that transcribes on-device and lets you point summaries at any model you want
LocalLLaMA
beerbellyman4vr70
🟧 echo.github ⭐The earliest primary artifact is the FastRepl GitHub repository, created with an initial commit by yujonglee on 2024-12-09 (then named Fastnyujonglee (FastRepl)——

Interpretation history

Decision trace