# BusinessKit Agent Your agents are useless without a knowledge layer. This is the knowledge layer — plus the agents that read from it. 17 agents (CEO, CRM, Social, SEO, Blog Writer, and more) that read and write directly to your Turso database. Two layers: **Brand Foundation (BF)** — static, you write once: `context/brand-voice.md`, `context/about-me.md`, `SOUL.md`. Agents read before producing anything. **Knowledge Base Layer (KBL)** — dynamic, agents maintain: `agent_kb` table grows every session. Drop anything into `agent_notes` inbox — links, ideas, tweets, articles. Run `/ingest` and the agent compiles structured wiki pages with cross-links. Every question you ask gets filed back as a new entry. Knowledge compounds. Works with Claude Code, Claude Cowork, Gemini CLI, Codex, and any terminal. --- ## Install — two ways ### Option A — Claude Code plugin (recommended) One command. Works across all projects. Auto-updates. Open Claude Code terminal (`Ctrl+`` ` ``): ```bash /plugin marketplace add businesskitai/businesskit-agent /plugin install businesskit@businesskitai /reload-plugins ``` Enter your credentials when prompted — `TURSO_URL` and `TURSO_TOKEN` from BusinessKit → Settings → Credentials. Commands are namespaced: `/businesskit:ceo`, `/businesskit:crm`, `/businesskit:blog-writer` etc. --- ### Option B — Clone for full access Source code, all agents, customise everything. ```bash git clone https://github.com/businesskitai/businesskit-agent cd businesskit-agent npm install cp .env.example .env # Paste TURSO_URL and TURSO_TOKEN from BusinessKit → Settings → Credentials npm run setup ``` Commands are direct: `/ceo`, `/crm`, `/blog-writer` etc. --- **Get credentials:** BusinessKit → Settings → Credentials → copy Turso URL and Token. --- ## Fill in your context (20 minutes, once) These files make every agent 10x better. Without them, agents guess. ``` context/about-me.md → who you are, your background, credibility context/brand-voice.md → voice bible with real writing examples context/working-style.md → how you work, approval preferences, schedule context/business.md → goals, revenue model, 90-day targets ``` **This is the most important step.** Shann Holmberg (Espressio) calls this the Brand Foundation — the static layer agents read before producing anything so the output sounds like you, not generic AI. --- ## How to use ### Claude Code Open this folder. Type any slash command: ``` /ceo → weekly business briefing, priorities, pending approvals /blog-writer → write posts, listicles, how-tos, comparisons /newsletter → draft and send newsletters /social → post to Twitter, LinkedIn, Instagram, TikTok (13 platforms) /crm → manage contacts, deals, outreach pipeline /seo → audit, fix meta, track LLM visibility /analytics → revenue, traffic, product performance /store → launch products, courses, services /ingest → process agent_notes inbox into knowledge base /kb → query your knowledge base /compress → save session before context runs out /resume → reload context next session ``` ### Claude Cowork Install the BusinessKit plugin. Same commands, runs in your browser. ### Terminal ```bash npx tsx cli.ts ceo npx tsx cli.ts blog-writer npx tsx cli.ts crm ``` --- ## The knowledge layer Drop anything into your inbox: ``` "add to notes: https://twitter.com/karpathy/status/... — useful for SEO strategy" "add to notes: idea — launch a free email course as lead magnet" "add to notes: competitor just raised prices, opportunity" ``` Run `/ingest` weekly. The agent reads each note, builds or updates wiki pages, cross-references related entries, marks contradictions as stale. Query anytime: ``` /kb → "what do I know about email deliverability?" "what content formats performed best?" "what's my current positioning vs competitors?" ``` Every answer gets filed back. The knowledge base gets richer every session. --- ## Agents | Agent | What they do | |---|---| | **CEO** | Weekly briefing, revenue pulse, priorities, pending approvals | | **CMO** | Marketing strategy, content calendar, growth planning | | **COO** | Operations, publishing queue, task management | | **CBO** | Revenue analysis, pricing strategy, product performance | | **Blog Writer** | Posts, listicles, how-tos, comparisons, programmatic content | | **Newsletter Writer** | Drafts, subscriber analytics, send management | | **Copywriter** | Landing pages, product copy, profile bio | | **Course Creator** | Course structure, lessons, pricing, access | | **Store Manager** | Products, webinars, services, sponsorships, 1:1 calls | | **Jobs Manager** | Job listings, applications | | **Forms Builder** | Intake forms, waitlists, surveys | | **Docs Writer** | Knowledge base articles, FAQs | | **CRM Agent** | Contacts, deals, outreach pipeline with approval gates | | **Social Agent** | 13 platforms via Zernio, scheduling, inbox | | **Analytics Agent** | Revenue, traffic, product, email stats | | **SEO Agent** | Technical audit, meta, LLM visibility tracking | | **Scheduler** | Cron tasks, publish queue, Hermes pattern detection | --- ## Cost Every agent is built to minimise token cost: - Skills loaded on demand — only what the task needs - KB index scanned first (slug + summary), content loaded only when needed - Reports pushed to Turso mid-conversation — never lost if session ends - `SYSTEM.md` enforces: do first, report after. No narrating what's about to happen. One founder moved from $130/5 days on a heavyweight agent setup to $10 on a lean knowledge-layer-first approach. That's the target. --- ## Works with | Tool | How | |---|---| | Claude Code | Primary runtime — slash commands in `.claude/commands/` | | Claude Cowork | Plugin at `.claude-plugin/plugin.json` — 18 commands | | Gemini CLI | Reads `GEMINI.md` automatically | | Codex | Reads `AGENTS.md` automatically | | Cursor | Reads `.cursorrules` automatically | | Any terminal | `npx tsx cli.ts ` | --- ## Requirements - Node.js 18+ - A BusinessKit account with Turso connected - Social platform tokens (optional — for `/social`) MIT License