> ## Documentation Index > Fetch the complete documentation index at: https://docs.augent.app/llms.txt > Use this file to discover all available pages before exploring further. # Workflows > Daily routines and practical workflows for Obsidian + Claude Code + augent. Your Obsidian vault is not a notebook. It is a searchable, linked, growing knowledge base. These workflows show how to treat it as one, with augent handling the heavy lifting: capturing content, finding connections, and turning raw research into finished writing. ## Capture: turning content into knowledge This is the highest-value workflow. You watch a talk, listen to a podcast, sit through a meeting. Without capture, it evaporates. With one command, it does not. ``` take_notes from https://youtube.com/watch?v=abc123 ``` Augent downloads the audio, transcribes it, generates structured notes, and saves everything to your vault. The output file includes YAML frontmatter: ```yaml theme={null} --- title: "Building Reliable AI Agents" tags: [AI, agents, reliability] date: 2026-03-23 type: notes source_url: https://youtube.com/watch?v=abc123 duration: "42:18" source_transcription: "[[Building Reliable AI Agents (transcription)]]" --- ``` The note body contains headers, key points, and timestamps. Wikilinks to related transcriptions in your vault are computed automatically based on semantic similarity. One command, and the content lives in your graph forever. ## Search: finding what you already know After three months of consistent capture, you have 50+ hours of transcribed content. The value is in retrieval. **Keyword search** finds exact mentions across every file: ``` search_memory "tokenomics" ``` Returns every transcription and note where the word appears, with surrounding context. **Semantic search** finds conceptually related content even when the exact words differ: ``` search_memory "scaling challenges for small teams" --mode semantic ``` This surfaces discussions about growth bottlenecks, hiring difficulties, and infrastructure limits, even if nobody said "scaling challenges" verbatim. **Deep search** drills into a single file for granular results: ``` deep_search "revenue model" in "Startup Fundraising Panel (transcription)" ``` Use keyword mode for precision. Use semantic mode for discovery. Use deep search when you know which file holds the answer. ## Review: daily briefings from your vault Claude can read your vault directly. Combine that with augent's search tools for a daily review routine. Ask Claude: "Read my last 5 transcriptions and tell me what themes keep appearing." It will pull the files, cross-reference topics, and surface patterns you missed. Or try: "What have I researched this week that I haven't written about yet?" Claude compares your recent captures against your published notes and drafts, flagging gaps. This works because every transcription is a structured markdown file with frontmatter, tags, and wikilinks. Claude does not need special access. It reads the files like you would, just faster. ## Create: writing from your own research Your vault is full of your thinking, captured in your voice, from sources you chose. Use it as raw material. Ask Claude: "Draft a thread about AI agents based on my notes from the last month." It pulls from `search_memory`, reads the relevant transcriptions, and writes using YOUR frameworks and YOUR examples. Or: "Find the strongest arguments across my research on decentralized identity and outline a blog post." Claude searches, ranks, and structures. You edit and publish. The difference between this and asking Claude to write from scratch: every claim traces back to something you actually watched, read, or discussed. ## Organize: keeping the graph healthy As your vault grows past a hundred files, structure matters. **Tagging** categorizes transcriptions with 2-4 broad topics each: ``` tag "Building Reliable AI Agents (transcription)" ``` Augent assigns semantic tags automatically based on the content. **Rebuilding the graph** recomputes all wikilinks and generates MOC (Map of Content) hub files: ``` rebuild_graph ``` MOC files act as cluster centers in Obsidian's graph view. Open the graph after a rebuild and you will see clear topic clusters: AI research here, product strategy there, technical deep-dives in another corner. Run `rebuild_graph` weekly, or after a batch of new captures, to keep connections fresh. ## Next steps Full list of augent tools with parameters and examples. Configure your Obsidian vault to work with augent's memory directory. Sync setup for Claude Code and Obsidian.