--- title: "Split one AI context file into a knowledge base your assistant can navigate" description: "When one CLAUDE.md gets too long, move to domain folders, context files, project files, templates, and decision logs with a stated read order." canonical: "https://scalewithsearch.com/articles/how-to-build-ai-knowledge-base" date: "2026-01-28" 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) # Split one AI context file into a knowledge base your assistant can navigate. You started with one context file, `CLAUDE.md`, about 300 words long. It said who you are, what you do, and how you work. It worked: the assistant stopped asking who you are, and conversations got more useful. Now you have several clients, different projects, and more workflows. The single file is crowded. You scroll to find the right section, and facts get buried. You do not need a bigger file. You need a structure of several files with a clear read order. This article shows when to split, how to organize the folders, and what each file holds, with a synthetic consultant as the example. ## Decide when to expand beyond one file One context file is enough when three conditions hold: - You work in one domain, such as client work only or content only. - You have no more than 3 to 5 active projects. - Your context is stable and does not change every week. Move to a multi-file structure when one or more of these is true: - You work in several domains, such as client work, internal operations, and personal projects. - You manage 5 or more clients or projects at the same time. - You need history: past decisions, completed projects, and lessons learned. - Other people need to read the knowledge base. - The single file is longer than 2,000 words and you cannot find facts in it. The practical signal is scrolling. When you scroll through the context file to find a fact, split it. ## Organize by domain, not by file type Do not organize by file type, with all emails in one folder and all templates in another. Organize by domain, the distinct areas of your work. Example domains are client work, content production, marketing and outreach, internal operations, product development, and personal projects. Each domain gets its own folder. Each folder holds everything for that domain: context, active projects, workflows, templates, and decision logs. The [guide to separate business contexts](/articles/separate-business-contexts-ai-agents) explains why a clean domain boundary also stops one client's facts from reaching another client's work. ## Use a folder structure that scales ```text /knowledge-base/ ├── CLAUDE.md (root context, read first) ├── client-work/ │ ├── _context.md (domain overview) │ ├── _active-projects.md │ ├── decision-log.md │ ├── client-a/ │ │ ├── project-context.md │ │ ├── deliverables.md │ │ └── meeting-notes.md │ ├── client-b/ │ │ ├── project-context.md │ │ └── deliverables.md │ └── templates/ │ ├── proposal-template.md │ └── status-update-template.md ├── content-production/ │ ├── _context.md │ ├── content-calendar.md │ ├── seo-strategy.md │ └── drafts/ ├── marketing/ │ ├── _context.md │ ├── outreach-campaigns.md │ └── email-templates/ └── operations/ ├── _context.md ├── workflows.md └── tools.md ``` ## Keep the root file short `CLAUDE.md` stays at the root. Claude Code loads it at the start of each session. It holds four things: - core identity: who you are and what you do; - voice rules: how the assistant writes for you; - a domain index: one line for each domain folder; - current priorities: the focus for this week or month. A synthetic root file looks like this: ```markdown # AI context ## Identity I am a fractional CMO for B2B SaaS companies. I help them move from founder-led sales to repeatable marketing systems. ## Voice - Direct, conversational, no jargon - Lead with the main point - Bullet lists, active voice, short sentences - Banned: solutions, leverage, ecosystem ## Domains - client-work/: active client projects, proposals, deliverables - content-production/: blog posts, newsletters, SEO content - marketing/: outreach campaigns, email sequences - operations/: internal workflows, tools, SOPs ## Read order For a client task, read client-work/_context.md, then the client folder, then the template the task names. ## Current focus (January 2026) - Client A: content strategy deliverable, due Feb 28 - Client B: paid acquisition work - Internal: course launch prep, March 15 ``` Keep this file under 500 words. It is the overview. Details live in the domain folders. ## Write a context file for each domain Each domain folder has a `_context.md` file. It states what the domain covers, the active projects, the standard workflows, and the key contacts or resources. ```markdown # Client work context ## Overview All paid client engagements. Current focus: content strategy and SEO for SaaS companies. ## Active clients - Client A (CRM for real estate teams): content and SEO - Client B (supply chain platform): paid acquisition - Client C (data visualization tool): onboarding ## Standard deliverables - Monthly status report, first Monday of each month - Quarterly strategy review, mid-quarter - Weekly Slack check-in, Fridays ## Pricing Quote only from rate-card.md. Do not state a price that is not in that file. ## Tools - Project tracking: Notion - Communication: Slack for clients, email for prospects - Deliverables: Google Docs, exported to PDF ``` The pricing rule is a pointer, not a number. When a price changes, you change one file. ## Give each project its own folder For each significant client or project, create a folder with three files. `project-context.md` holds background, goals, timeline, and contacts: ```markdown # Client A: project context ## Client background - Product: CRM for real estate agents - Stage: Series A, 40 employees - Market: US real estate teams of 5 to 20 agents ## Engagement - Start date: 2026-01-15 - Contract: 6-month retainer (terms in the signed agreement) - Primary contact: VP Marketing (name and email in the CRM) ## Goals 1. Increase organic traffic by 150% from a baseline of 8,000 visits a month 2. Rank for 20 high-intent keywords (category and use-case terms) 3. Build a content engine that produces 8 articles a month ## Timeline - Months 1 to 2: audit and strategy - Months 3 to 6: execution and optimization - Next deliverable: Feb 28, topical authority map and content calendar ``` `deliverables.md` holds status and due dates: ```markdown # Client A: deliverables ## Due Feb 28, 2026 - [ ] Topical authority map (draft done, review pending) - [ ] Q1 content calendar (in progress) - [x] Keyword research document ## Due March 31, 2026 - [ ] First 8 articles published - [ ] Internal linking structure in place - [ ] Author bio pages ## Completed - [x] SEO audit, delivered Jan 22 - [x] Competitive analysis, delivered Jan 29 - [x] Keyword opportunity document, delivered Feb 5 ``` `meeting-notes.md` holds discussions, decisions, and action items: ```markdown # Client A: meeting notes ## 2026-01-22: kickoff call Attendees: Client A VP Marketing, consultant Decisions: - Start with bottom-of-funnel keywords (comparisons, use cases) - Audience: real estate team leads, not solo agents - Tone: professional, approachable, heavy on case studies Action items: - Consultant: deliver audit by Jan 29 (done) - Client: grant Google Analytics and Search Console access (done) - Client: share five competitor sites (done) Next meeting: Feb 12, 2026 ``` ## Store templates beside the work Keep reusable templates in each domain folder. When you ask for a draft, the assistant fills the template. ```markdown # Proposal template ## [Client name]: [service] proposal Prepared for: [contact name and title] Prepared by: [your name, your company] Date: [YYYY-MM-DD] ### The challenge [Two or three sentences on the client's problem] ### The approach [What you will do, at a high level] ### Scope of work Phase 1: [name] (weeks 1 to 2) - [Deliverable 1] - [Deliverable 2] Phase 2: [name] (weeks 3 to 8) - [Deliverable 3] - [Deliverable 4] ### Timeline - Kickoff: [date] - Phase 1 complete: [date] - Phase 2 complete: [date] ### Investment - [Pricing structure from rate-card.md] - Payment terms: [details] ### Next steps [What happens if the client says yes] ``` ## Record decisions in a log You make decisions as you work: why you chose one approach over another, why you recommended a strategy, why you turned down a project. Write them down. You will need them later, and so will the assistant. ```markdown # Client work: decision log ## 2026-01-15: bottom-of-funnel content first Context: Client A wanted to rank for broad terms such as "real estate CRM." Decision: start with comparisons, use cases, and alternatives. Reasons: - Broad terms are highly competitive and would take 12 months or more - Bottom-of-funnel terms have higher purchase intent - They rank sooner and build domain authority - Broad terms can follow in months 6 to 12 Result: client approved the strategy. ## 2026-01-10: declined a construction tech prospect Context: inbound lead, construction project software, budget above the usual range. Decision: declined. Reasons: - No construction tech experience; 20 or more hours of research needed - Six-week launch timeline was not realistic - Budget did not cover the learning curve - Current load at capacity with 3 active clients Lesson: do not take work outside core expertise because the budget is high. ``` The [AI correction log template](/articles/ai-correction-log-template) uses the same pattern for mistakes the assistant must not repeat. ## Tell the assistant how to read the knowledge base Claude Code loads `CLAUDE.md` automatically. It opens other files when the root file tells it where to look or when you name a file. It does not guess a folder structure reliably, so write the read order into the root file. With the read order above, the request "Draft a status update for Client A" leads the assistant to five files: 1. `CLAUDE.md`, for identity and voice rules; 2. `client-work/_context.md`, for the client workflow and report format; 3. `client-work/client-a/project-context.md`, for the client, timeline, and goals; 4. `client-work/client-a/deliverables.md`, for current status; 5. `client-work/templates/status-update-template.md`, for the structure. The draft has the correct client name, accurate status, your format, and your voice. Ask the assistant to list the files it read at the end of the draft. That list is your check that the read order works. For a recurring job, a [context manifest](/articles/ai-context-manifest-template) fixes the source list in one file. ## Give each fact one home A split creates a new failure: the same fact in two files. Suppose the root file says "Client A report due on the first Monday." Suppose `client-work/_context.md` says "reports go out on the fifth business day." The assistant reads both and picks one, and you cannot predict which. Give each fact one home, and point to it from every other file. Put the report schedule in `client-work/_context.md`. In the root file, write "Report schedule: see client-work/_context.md." A pointer cannot go stale, because it holds no value. Run a duplicate check after each split. Pick five facts that change often: a deadline, a rate, a contact role, a status, and a tool. Search the knowledge base for each one. If a value appears in two files, delete one copy and put a pointer in its place. ## Test the split with a cold session Open a new session. Ask three questions whose answers sit in different folders: 1. "When is the next Client A deliverable due?" The answer is in `client-work/client-a/deliverables.md`. 2. "Which template do I use for a status update?" The answer is in `client-work/templates/`. 3. "Why did I decline the construction tech prospect?" The answer is in `client-work/decision-log.md`. For each answer, check three things: the value, the source file the assistant names, and the absence of an older value. A wrong source file means the read order needs one more line. A missing answer means the fact has no home yet. Add the line or the fact, then run the test again in a new session. ## Keep the knowledge base current The system works only when it stays current. Set a rhythm: | Interval | Task | |---|---| | Daily | Add meeting notes and update deliverable status | | Weekly | Review active projects and update priorities in the root file | | Monthly | Archive completed projects, update domain context files, review decision logs | Most of this is work you already do. You write it in a place the assistant can read. ## Choose an editor Obsidian is built for linked Markdown files. You can connect notes, search the whole vault, and see links between files. Obsidian is free to download. The notes stay on your disk, but the text Claude Code reads from them goes to Anthropic for processing. The [Obsidian business-memory guide](/articles/obsidian-ai-memory-for-business) covers the vault layout. VS Code works well if you already use it for code. It has good search, folder management, and Markdown preview. Notion exports to Markdown, but native Notion databases do not suit direct file reading by an assistant. Organize in Notion if you must, then export for the assistant. The [owned business memory and Notion AI comparison](/articles/owned-business-memory-vs-notion-ai) lists what a Notion export keeps and what it loses. Plain folders and a text editor such as Notepad or TextEdit also work. If you can organize folders and write Markdown, you have enough. ## Grow the system in stages Do not build everything at once. 1. Start with one `CLAUDE.md` file of 300 to 500 words. 2. Add domain folders, each with a `_context.md` file. 3. Add project folders for active clients and projects. 4. Add templates, decision logs, and workflows. 5. Archive completed projects to build a historical reference. Move to the next stage when the current one causes friction. With the full structure in place, you switch between projects without re-explaining. A new team member can read the knowledge base to learn the business. Every decision, project, and workflow has a searchable record, and each new document gives the assistant more to work from. ## Related: AI memory - [Build a local-first AI setup that sends only the context each request needs](/articles/local-first-ai-architecture) - [Set up an Obsidian vault that routes Claude Code to the right context](/articles/obsidian-ai-vault-setup-guide) - [Turn your second brain into context an AI reads every session](/articles/second-brain-for-ai) ---- ```text .|########||. .|########||. .|########||. |##||. .||##|. |##||. .||##|. |##||. .||##|. |#|. .|#|. |#|. .|#|. |#|. .|#|. |#| |#| |#| |#| |#| |#| .#| |#. .#| |#. .#| |#. .#. .#| .#. .#| .#. .#| |#. .#| |#. .#| |#. .#| |# ...... #| |# ...... #| |# ...... #| .# ||#########| #| .# ||#########| #| .# ||#########| #| .#. |######||######|. .#. .#. |######||######|. .#. .#. |######||######|. .#. #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# # .||####|###########################|. # # .||####|###########################|. # # .||####|###########################|. # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || #. .######|#|#########| .# #. .######|#|#########| .# #. .######|#|#########| .# .#. |#####||||#####| .#. .#. |#####||||#####| .#. .#. |#####||||#####| .#. |# ||########|| #. |# ||########|| #. |# ||########|| #. |# ...... #| |# ...... #| |# ...... #| |#. .#| |#. .#| |#. .#| |#. .#. |#. .#. |#. .#. .#| |#. .#| |#. .#| |#. |#| |#| |#| |#| |#| |#| .|#|. .|#|. .|#|. .|#|. .|#|. .|#|. |##||. .||##| |##||. .||##| |##||. .||##| .||########||. .||########||. .||########||. Scale With Search 2026 [scalewithsearch.com](https://scalewithsearch.com) ```