--- title: "Build your AI workflow beyond the first CLAUDE.md." description: "After Claude Code init, map your retrieval, automation, alerts, and outside-system gates. Follow the dependency chain before adding each layer." canonical: "https://scalewithsearch.com/articles/beyond-claude-code-init" date: "2026-03-23" modified: "2026-09-25" --- ## Site navigation - [Scale With Search](https://scalewithsearch.com/) - Real estate - Real estate - [Real estate](https://scalewithsearch.com/for/real-estate) - Work - Start here - [Send your brief](https://scalewithsearch.com/work#send-your-brief) - [Prepare your six-question brief](https://scalewithsearch.com/work#prepare-your-six-question-brief) - Build - [Site build, content library with SEO, signal desk](https://scalewithsearch.com/work) - For your business - Trades and home services - [Auto body and collision shops](https://scalewithsearch.com/for/auto-body-and-collision-shops) - [Foundation and home repair contractors](https://scalewithsearch.com/for/foundation-and-home-repair) - [Garage door and fencing contractors](https://scalewithsearch.com/for/garage-door-and-fencing-contractors) - [HVAC contractors](https://scalewithsearch.com/for/hvac-contractors) - [Janitorial and commercial cleaning companies](https://scalewithsearch.com/for/janitorial-and-commercial-cleaning) - [Locksmiths](https://scalewithsearch.com/for/locksmiths) - [Moving companies](https://scalewithsearch.com/for/moving-companies) - [Pest control companies](https://scalewithsearch.com/for/pest-control-companies) - [Plumbing and electrical contractors](https://scalewithsearch.com/for/plumbing-and-electrical-contractors) - [Restoration and water or fire damage companies](https://scalewithsearch.com/for/restoration-and-water-fire-damage) - [Roofing companies](https://scalewithsearch.com/for/roofing-companies) - [Towing companies](https://scalewithsearch.com/for/towing-companies) - [Tree services and landscaping companies](https://scalewithsearch.com/for/tree-services-and-landscaping) - [Solar installers](https://scalewithsearch.com/for/solar-installation) - [General contractors](https://scalewithsearch.com/for/general-contractors-and-construction) - [Paving, concrete, and flooring contractors](https://scalewithsearch.com/for/paving) - Practices and professional services - [Bookkeeping and tax practices](https://scalewithsearch.com/for/bookkeeping-and-tax-practices) - [Dental practices](https://scalewithsearch.com/for/dental-practices) - [Family and criminal defense law firms](https://scalewithsearch.com/for/family-and-criminal-defense-law-firms) - [Med spas and aesthetics practices](https://scalewithsearch.com/for/med-spas-and-aesthetics) - [Personal injury law firms](https://scalewithsearch.com/for/personal-injury-law-firms) - [Veterinary clinics](https://scalewithsearch.com/for/veterinary-clinics) - [Gyms and fitness studios](https://scalewithsearch.com/for/fitness) - [Therapy and outpatient health practices](https://scalewithsearch.com/for/therapy-and-outpatient-health) - [Medical billing companies](https://scalewithsearch.com/for/medical-billing) - [Insurance agencies](https://scalewithsearch.com/for/insurance-agencies) - [Financial advisors](https://scalewithsearch.com/for/financial-advisors) - [Property management companies](https://scalewithsearch.com/for/property-management) - [Recruiting and staffing agencies](https://scalewithsearch.com/for/recruiting-and-staffing) - [Architects and interior designers](https://scalewithsearch.com/for/architects-and-interior-designers) - [Logistics and supply chain companies](https://scalewithsearch.com/for/logistics-and-supply-chain) - Agencies, MSPs, and manufacturing - [IT and managed service providers](https://scalewithsearch.com/for/it-and-managed-service-providers) - [Machine shops and precision manufacturers](https://scalewithsearch.com/for/machine-shops-and-precision-manufacturing) - [Marketing agencies and freelancers](https://scalewithsearch.com/for/marketing-agencies-and-freelancers) - [SEO agencies and consultants](https://scalewithsearch.com/for/seo-agencies-and-consultants) - [Small manufacturers and fabricators](https://scalewithsearch.com/for/small-manufacturers-and-fabricators) - Restaurants, shops, studios, and nonprofits - [Restaurants and hospitality businesses](https://scalewithsearch.com/for/restaurants-and-hospitality) - [Retail stores and ecommerce sellers](https://scalewithsearch.com/for/retail-and-ecommerce) - [Photographers, event planners, and travel agents](https://scalewithsearch.com/for/photographers) - [Churches and nonprofits](https://scalewithsearch.com/for/churches-and-nonprofits) - [All industries](https://scalewithsearch.com/for/) - Learn - For your office - [Office job guides](https://scalewithsearch.com/guides/) - [Browser calculators](https://scalewithsearch.com/tools/) - Start here - [How it works](https://scalewithsearch.com/how-it-works) - [Free Starter Kit](https://scalewithsearch.com/kit/business-memory-starter-kit.zip) - [Synthetic specimen](https://scalewithsearch.com/specimen/working-session-specimen.zip) - Guides - [The Complete Guide to Business Memory for AI Agents](https://scalewithsearch.com/articles/business-memory-for-ai-agents-guide) - [The Complete Small-Business Guide to AI Agent Governance](https://scalewithsearch.com/articles/ai-agent-governance-guide-small-business) - [The Complete Guide to Leaving Vendor AI Memory](https://scalewithsearch.com/articles/leaving-vendor-ai-memory-guide) - Articles by cluster - [Business memory](https://scalewithsearch.com/articles/business-memory-for-ai-agents-guide) - [Agent governance](https://scalewithsearch.com/articles/ai-agent-governance-guide-small-business) - [Migration and ownership](https://scalewithsearch.com/articles/leaving-vendor-ai-memory-guide) - For machines - [llms.txt](https://scalewithsearch.com/llms.txt) - [llms-full.txt](https://scalewithsearch.com/llms-full.txt) - [Machine view](https://scalewithsearch.com/?view=machine) - Company - Evidence - [Proof](https://scalewithsearch.com/proof) - Company - [About](https://scalewithsearch.com/about) # Map the four layers a production AI memory system adds after /init. You ran `/init`. Claude Code scanned the repository, wrote a `CLAUDE.md`, and now reads that file at the start of each session. It no longer asks which language the project uses. That solves the cold start for one project in one tool. Most tutorials end there. This page describes what comes next, based on one production system as it ran in March 2026. That system governed more than 3,000 files across five business domains. It could recall conversations from six months earlier. It sent phone alerts at 7 a.m. It refused to write to the CRM without explicit approval. The system has 14 parts in four layers. `CLAUDE.md` is the first part. Each layer depends on the one below it: data feeds automation, automation gives communication something to report, and communication makes integration safe. If you skip a layer, the layers above it fail. ## Know what /init produces and where it stops The `/init` command examines a codebase and writes a Markdown file that Claude Code loads at session start. The file usually records the language, framework, test commands, lint rules, and directory structure. For a code project, it is accurate enough to work. The file describes the project. It does not describe the operator or the business. It holds no communication style, business rules, or decision frameworks. It lacks the hard constraints that an owner refines over hundreds of sessions. The system on this page carried 19 of them. It also lacks the names and roles of the people the agent deals with. A project configuration file and a persistent business record solve different problems. The generated file also cannot reach outside knowledge. It does not know what sits in your notes, your CRM, or your session history. It describes one directory. In a production system, `CLAUDE.md` is the entry point to a network of sources. The [context file and CLAUDE.md guide](/articles/context-file-vs-claude-md-business-project) shows which content belongs in the entry point and which belongs in separate files. ## Layer 1: make the knowledge findable `CLAUDE.md` tells the agent what the project is. The data layer gives it access to what the business knows. Three systems make up this layer. **System 2: vault structure.** An Obsidian vault holds one directory per domain. In the example system, the domains covered sales operations, an SEO practice, course builds, personal finance, and contacts. Each domain has a `_context.md` file. A routing table in `CLAUDE.md` maps keywords to domains. A mention of the CRM or a lead list loads the sales context. A mention of a topical map or a retainer loads the SEO context. Five domains and five context files leave no doubt about which role the agent plays. The [guide to separate business contexts](/articles/separate-business-contexts-ai-agents) covers the isolation rules that keep the domains apart. **System 3: semantic search.** `grep` finds exact strings. It fails when you need "the note about the client who wanted acreage outside town" and you remember no filename, folder, or unique keyword. The example system used QMD, which indexes every vault file as vector embeddings. The agent then searches by meaning, not by text match. Across 3,000 files, this is the difference between a filing cabinet and a usable memory. **System 4: session ledger.** A SQLite database with full-text search held more than 37,000 messages: every conversation, instruction, and correction. When the agent needs a decision from three months ago about a client's pricing structure, it queries the ledger. The ledger survives session ends, context window resets, and model upgrades. The [full-text chat history search guide](/articles/search-claude-chat-history-full-text) covers a smaller version of the same idea. With these three systems, the agent can find any file by meaning, recall a past conversation by keyword, and load the right domain context. Every layer above depends on these three retrieval paths. ## Layer 2: let work run without you Data gives the agent knowledge. Automation gives it work to do while you are away. Four systems make up this layer. **System 5: hooks.** Claude Code hooks run on specific events and can add context before the model processes a prompt. Mention a client name, and a hook loads that client's contact file. Mention a deal, and a hook loads the transaction record. The agent does not choose to search. The pipeline is deterministic: keyword detected, context loaded, no missed lookup. **System 6: skills.** A skill packs a repeatable workflow into one command. In the example system, `/sprint` loaded the current sprint and checked the backlog. `/handoff` wrote a session summary for the next agent. `/deals` ran a dry-run audit of CRM data under seven safety constraints. Each skill replaced about five minutes of prompts, and more than 20 skills replaced more than 20 prompt templates. **System 7: timers and scheduled tasks.** `systemd` timers on a VPS run tasks on a schedule. The example system captured podcast transcripts daily at 6 a.m. It sent a sprint reminder at 7 p.m. on Wednesdays. At midnight, it checked that each integration endpoint responded and that API tokens had not expired. You wake up to results, not to a task list. **System 8: proactive engine.** One orchestration layer sits above the individual timers. Instead of 22 separate cron jobs with 22 separate failure modes, one engine dispatches tasks, watches for completion, retries, and reports status. When a podcast source changes its URL structure, the engine catches the 404 error, logs it, and sends an alert. A timer that fails without output does not go unnoticed. ## Layer 3: reach the owner away from the terminal An agent that runs tasks overnight is of little use if you do not know what it did. Three systems make up this layer. **System 9: domain bots.** The example system ran five Telegram bots, one per domain. The sales bot sent lead alerts. The personal bot sent household reminders. The SEO bot sent client notices. Because each domain has its own bot, you can mute one during off-hours and keep the others. Each bot has its own token, message format, and batch rules. The sales bot batched alerts into one morning digest; the personal bot sent at once. **System 10: notification routes.** Routes set quiet hours, priority levels, and channels. A failed health check at 3 a.m. does not wake you unless it affects critical infrastructure. A new lead at 7:30 a.m. arrives through the sales bot with the CRM record ID embedded, so one tap opens the record. Everything else waits for its scheduled time. **System 11: health monitors.** Endpoint checks, token expiry tracking, and service status checks close the loop. When the CRM API returns a 401 error, you know within minutes. Without the monitor, you learn three days later, when a client asks about a missed follow-up. The system does not only run tasks. It confirms that they succeeded and flags the ones that did not. The agent works in a terminal. The owner works from a phone, a car, and an office. This layer carries status between them without a laptop and a log file. ## Layer 4: touch outside systems under gates The first three layers are internal. The fourth layer is where the agent acts on systems that affect other people. Three systems make up this layer. **System 12: CRM integration with safety gates.** The agent can read from the CRM: contacts, deal stages, and response history. By default, it cannot write. Every CRM write runs in dry-run mode first. Seven rules govern what the agent may change, under which conditions, and with which approval. One bad batch update to 200 contact records takes hours to undo. A dry-run default keeps the agent's analysis useful and its reach into live records at zero until someone lifts the gate. **System 13: outreach automation.** Outreach runs in three tiers: relationship, presence, and opportunity. Each tier has its own approval step. The agent drafts the message, selects recipients by pipeline stage, and proposes timing from engagement patterns. It does not send. Each outreach action waits in a queue for review. A governing document sets tone, frequency caps, and exclusion rules. The [approval guide for agent sends](/articles/who-approves-what-an-ai-agent-sends) describes how to bind an approval to the exact message. **System 14: content schedules.** Article pipelines, social queues, and newsletter drafts draw on vault context, session history, and domain knowledge from all three lower layers. The schedule sets publication times across platforms. The content reflects the owner's actual positions because the agent reads the full knowledge base, not a one-paragraph bio. The gates exist because these actions cannot be undone the way a file edit can. ## Trace the dependency chain Each layer needs the one beneath it. Automation without data gives scheduled tasks that cannot find what they need. A timer fires to check a deal's status, but no semantic search locates the deal file and no ledger recalls the last conversation about it. Communication without automation means manual status checks. You open the terminal, run the query, and read the output. The bot has nothing to report, because nothing ran on its own. Integration without communication means blind writes. The agent changes CRM records, sends outreach, or publishes content. You learn about it when a client replies to a message you did not approve. `CLAUDE.md` is the root. Remove it, and Claude Code does not know your rules, domains, voice, or constraints. Hooks do not know which keywords to catch. Skills do not know which gates to enforce. Bots do not know their domain. That is why `/init` matters, and why it is the first of fourteen steps. ## See one morning run through the stack On a Monday morning, before the owner leaves for the office, the phone shows three messages. The sales bot batched overnight activity: two new inquiries, and one hot prospect with a 4-day follow-up gap. The SEO bot confirmed that a client's weekly article went live on schedule. The health monitor reported that every API token stays valid for another 30 days. At 8 a.m., the owner opens Claude Code. It already knows the lead with the follow-up gap, because a hook loaded the contact file when the digest named the lead. The owner types "draft the follow-up." The agent writes the email in the owner's voice and names the property the lead asked about. It matches the tone of the last three messages to similar prospects, which it pulls from the ledger. It queues the draft for approval. One tap approves and sends it. That sequence touched 9 of the 14 systems. The owner typed three words. With only `/init`, the same task goes differently. You open Claude Code and explain who you are. You describe the lead, paste the property details, and specify the tone. You review, edit, and send by hand, then repeat it tomorrow. ## Build the layers in order Apart from the model, every system on this page runs on free or open-source tools: Obsidian, Telegram bots, SQLite, QMD, and `systemd` timers. Claude Code itself needs a paid Claude plan or an Anthropic API account. Start with the vault structure, the routing table, and the `CLAUDE.md` structure, because the rest depends on them. The [business CLAUDE.md template](/articles/claude-md-template-business-context) gives a starting file. In the example system, the vault and `CLAUDE.md` took one sitting with guidance, or a weekend for someone who learns alone. The full four-layer stack took two to three weeks of focused build time. Each layer took two to four days once the layer below it was stable. The order is fixed. You cannot wire the CRM gates before semantic search works, because the gates refer to vault files that must be findable. In March 2026, the only added running cost over a bare `CLAUDE.md` was a small VPS for timers, health checks, and the proactive engine. It cost less per month than the AI subscription itself. Obsidian is free to download, Telegram bots cost nothing, and QMD and the session ledger ran on the local machine. The index files stay on that disk, but the text the agent retrieves from them goes to the model provider for processing. The larger cost is the owner's build time and upkeep. ## Related: AI memory - [Install Obsidian and Claude Code, then connect your first AI vault](/articles/how-to-set-up-ai-vault) - [Turn your second brain into context an AI reads every session](/articles/second-brain-for-ai) - [Keep your context when an AI chatbot resets between sessions](/articles/why-do-ai-chatbots-forget) ---- ```text .|########||. .|########||. .|########||. |##||. .||##|. |##||. .||##|. |##||. .||##|. |#|. .|#|. |#|. .|#|. |#|. .|#|. |#| |#| |#| |#| |#| |#| .#| |#. .#| |#. .#| |#. .#. .#| .#. .#| .#. .#| |#. .#| |#. .#| |#. .#| |# ...... #| |# ...... #| |# ...... #| .# ||#########| #| .# ||#########| #| .# ||#########| #| .#. |######||######|. .#. .#. |######||######|. .#. .#. |######||######|. .#. #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# # .||####|###########################|. # # .||####|###########################|. # # .||####|###########################|. # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || #. .######|#|#########| .# #. .######|#|#########| .# #. .######|#|#########| .# .#. |#####||||#####| .#. .#. |#####||||#####| .#. .#. |#####||||#####| .#. |# ||########|| #. |# ||########|| #. |# ||########|| #. |# ...... #| |# ...... #| |# ...... #| |#. .#| |#. .#| |#. .#| |#. .#. |#. .#. |#. .#. .#| |#. .#| |#. .#| |#. |#| |#| |#| |#| |#| |#| .|#|. .|#|. .|#|. .|#|. .|#|. .|#|. |##||. .||##| |##||. .||##| |##||. .||##| .||########||. .||########||. .||########||. Scale With Search 2026 [scalewithsearch.com](https://scalewithsearch.com) ```