AI Company Starter Stack¶

What it is¶

The AI Company Starter Stack is an opinionated selection of tools and architectural patterns designed to turn a traditional business into an AI-native organization. It provides a curated list of "defaults" across various layers—from web surfaces and agent operating models to workflow control planes and local inference options—enabling teams to build operating leverage rather than just side experiments.

What problem it solves¶

Most organizations struggle with "tool sprawl" when adopting AI, often implementing fragmented solutions that don't communicate or scale. This starter stack solves that by providing a unified operating system where n8n coordinates workflows, Claude Skills package procedures, and mem0 preserves context. It focuses on the "smallest stack" that provides maximum leverage across product, operations, and research.

Where it fits in the stack¶

Category: Knowledge Base / Architectural Pattern. It serves as the operational blueprint of the repository, integrating various tools from the docs/services/ and docs/tools/ directories into a cohesive business framework.

Typical use cases¶

Bootstrapping an AI Agency: Using the "Research and lead-intel pack" to automate client research and market synthesis.
Internal Operations Modernization: Implementing the Google Workspace CLI and n8n to automate administrative drudgery.
Privacy-Conscious R&D: Setting up LocalAI and llmfit for secure, internal-only AI development.
Rapid MVP Launch: Following the "Website launch pack" to build and deploy a public-facing AI product shell.

Strengths¶

High Cohesion: Tools are selected based on how well they integrate with each other (e.g., n8n + Paperless + Vikunja).
Cost-Efficiency: Prioritizes free or low-cost starter tiers and local inference options.
Scalable: Provides a clear "Replace when" path for every layer, ensuring the stack grows with the company.
Outcome-Focused: Categorized into "Expansion packs" targeted at specific business results.

Limitations¶

Opinionated: The "Default choice" may not fit companies with strict legacy infrastructure constraints (e.g., non-Google Workspace environments).
Maintenance Overhead: Running a full self-hosted stack (n8n, Supabase, LocalAI) requires more technical expertise than using SaaS-only solutions.
Fast Obsolescence: The "Frontier" models and tools change rapidly, requiring periodic review of the recommended defaults.

When to use it¶

When you are building a new company and want to be "AI-native" from day one.
When your current AI efforts are fragmented and you need a unified operating model.
When you want to reduce per-token costs by migrating some workloads to local/private inference.

When not to use it¶

If you already have a mature, high-scale AI infrastructure that requires specialized, non-standardized tools.
If you are looking for a single-app solution rather than a comprehensive company-wide stack.

Getting started¶

To implement the AI Company Starter Stack:

Review the Default Stack: Scan the table below to understand the core layers and recommended tools.
Select an Expansion Pack: Choose the pack (e.g., "Knowledge workspace pack") that matches your most urgent business bottleneck.
Deploy the Core: Start with n8n and the Google Workspace CLI to build your first automation.
Iterate with Skills: Package your repeatable procedures into the Claude Skills Ecosystem to empower your team.

The default stack¶

Layer	Default choice	Why it belongs in the starter stack	Replace when
Website / app surface	Free AI Website Playbook -> Vercel + Supabase	Gives you a practical default for public sites, MVPs, and internal tools	Your architecture has stronger static-site or infra constraints
Agent operating model	Claude Code + MCP	Native terminal integration and extensible context	You need high-level visual workflow orchestration
Current technical context	Context7	Keeps coding agents grounded in current docs	Your work is mostly non-technical or repo-local
Workflow control plane	n8n	Scheduling, retries, approvals, and cross-system automation	Your company is still too small for durable workflows
Web interaction	Browser Use	Covers UI-only systems and interactive web tasks	Stable APIs exist for the same work
Company operating surface	Google Workspace CLI	Documents, calendars, files, and admin actions become automatable	You do not run the company on Google Workspace
Memory	mem0	Preserves account, process, and user context across sessions	Tasks are stateless and one-off
Research harness	DeerFlow	Strong fit for strategy, research, and evidence synthesis	Your work is mostly transactional automation
Claude implementation examples	Claude Cookbooks	First-party examples reduce integration guesswork	You already have strong internal patterns
Local/private inference option	LocalAI + llmfit	Gives you a path to self-hosting without blind hardware bets	Frontier cloud models are more important than control

Expansion packs¶

Pack	Add these tools	Use when	Main output
Website launch pack	Free AI Website Playbook + Vercel + Supabase	You need a public launch surface, waitlist, or early MVP	Public web presence and product shell
Knowledge workspace pack	AnythingLLM + LocalAI or Ollama	Teams need a shared internal assistant over documents and knowledge	Internal AI workspace
Routing and model control pack	LiteLLM + OpenRouter + ClawRouter	You need cost, fallback, or provider-routing control across many agents	Model governance and routing
Research and lead-intel pack	DeerFlow + Tavily + Browser Use + mem0	Sales, strategy, or market research becomes a core workflow	Reusable intelligence pipelines
Finance and market pack	OpenBB + n8n + Google Workspace CLI	Finance, investor, or market reporting needs become recurring	Structured finance briefings
Visual builder pack	Flowise + Supabase	You want non-engineers or mixed teams to ship internal AI tools faster	Prototype-to-internal-app path

What each part is for¶

0. Website layer¶

Use this when the company needs an actual web surface, not only internal automation.

The Free AI Website Playbook helps choose what kind of site to build and what host fits.
Vercel is the default when speed of launch matters.
Supabase is the default backend once forms, auth, or product state appear.
Use GitHub Pages for docs-heavy sites and Cloudflare Pages for static directories and static-first public properties.

1. Claude Code + MCP Servers¶

Use this pair when you want agents to behave like trained operators with direct access to your infrastructure and data.

Claude Code provides a high-leverage terminal interface for agentic coding and operations.
MCP (Model Context Protocol) provides standardized connectors to databases, APIs, and local filesystems.
Together, they make AI natively integrated into the developer and operator workflow.

2. n8n + Workspace CLI¶

Use this pair when the company needs actual operating workflows.

n8n coordinates timing, branching, retries, approvals, and logs.
Google Workspace CLI executes useful work in Docs, Sheets, Drive, Calendar, and Chat.
Together, they turn office operations into an automatable system.

3. Context7 + Claude Cookbooks¶

Use this pair when the team is building AI products or internal tools.

Context7 provides current third-party docs.
Claude Cookbooks provides first-party Claude implementation patterns.
Together, they reduce bad assumptions during build work.

4. Browser Use + mem0 + DeerFlow¶

Use this trio for research, lead generation, and web-native execution.

Browser Use interacts with websites.
mem0 remembers what happened and what matters.
DeerFlow structures longer research/execution tasks.

5. LocalAI + llmfit¶

Use this pair when privacy, local control, or cost discipline matters.

llmfit decides what can run on your hardware.
LocalAI gives you a reusable local API once you know the hardware plan is viable.

6. Routing layer: LiteLLM vs OpenRouter vs ClawRouter¶

Use this layer when model choice, fallback behavior, or provider economics become operational concerns.

LiteLLM is the best general internal routing/control plane.
OpenRouter is the fastest way to access many hosted models behind one API and billing layer.
ClawRouter is the specialized choice when OpenClaw-style agent routing is part of the architecture itself.

7. Knowledge workspace: AnythingLLM¶

Use this when the company needs a usable AI workspace for teams, not just back-end components.

AnythingLLM is a fast way to stand up internal knowledge assistants.
It is especially useful when the team wants document-grounded chat before investing in custom product work.

8. Finance intelligence: OpenBB¶

Use this when structured financial or market intelligence should feed founder, investor, or sales workflows.

OpenBB adds structured market data that generic web search cannot replace.
It is most useful for companies where market, macro, or financial context changes decisions.

Comparison table¶

Need	Best default	Use instead when	Comment
Public website or product shell	Free AI Website Playbook + Vercel	Static docs or content site is enough	The playbook is the decision layer; Vercel is the common default
Reusable company procedures	Claude Skills Ecosystem	You only need one-off prompts	Skills matter once the process repeats
Reliable coding-agent execution	Superpowers	Speed matters more than rigor	Best for important engineering work
Workflow orchestration	n8n	One small script is enough	n8n is the operating system, not just a node editor
Browser-only workflows	Browser Use	API exists	APIs beat browsers when available
Workspace automation	Google Workspace CLI	Your company runs elsewhere	Great when Docs/Sheets/Drive are core
Long-lived memory	mem0	Stateless execution is enough	Do not add memory unless it pays for itself
Deep research	DeerFlow	Search + one summary call is enough	Best for evidence-heavy work
Current docs for agents	Context7	No external docs required	Very high leverage for coding teams
First-party Claude patterns	Claude Cookbooks	You need broader ecosystem docs	Strong complement to Context7
Local inference	LocalAI	You need frontier quality and speed first	Pair with llmfit before buying hardware
Internal AI workspace	AnythingLLM	You need full custom product behavior	Great for shared internal assistants
Model routing	LiteLLM	You only need one hosted provider	Use OpenRouter for simpler hosted access; use ClawRouter for OpenClaw-heavy stacks
Finance / market intelligence	OpenBB	Generic web research is enough	Best when market data should be structured and recurring

Selection map¶

flowchart TD
    A["What is the main bottleneck?"] --> B["Execution quality"]
    A --> C["Workflow automation"]
    A --> D["Web-native tasks"]
    A --> E["Research and synthesis"]
    A --> F["Privacy / local control"]

    B --> B1["Skills + Superpowers + Context7 + Claude Cookbooks"]
    C --> C1["n8n + Google Workspace CLI + Supabase"]
    D --> D1["Browser Use + n8n"]
    E --> E1["DeerFlow + Tavily + mem0"]
    F --> F1["llmfit + LocalAI (+ Ollama when simpler local serving is enough)"]

Overlap map¶

flowchart LR
    S["Skills / Superpowers"] --- O["Operating discipline"]
    N["n8n / Workspace CLI"] --- O
    B["Browser Use"] --- W["Web execution"]
    D["DeerFlow"] --- R["Research"]
    M["mem0"] --- R
    M --- W
    C["Context7 / Claude Cookbooks"] --- O
    L["LocalAI / llmfit"] --- P["Private inference"]
    P --- O
    K["AnythingLLM"] --- O
    R2["LiteLLM / OpenRouter / ClawRouter"] --- O
    F2["OpenBB"] --- R

Example starter stacks¶

Lean AI-native services company¶

Use this when you want the smallest serious stack for delivery, documentation, and operations.

Research-heavy AI consultancy¶

Use this when account research, competitor analysis, and evidence gathering are core to the business.

Privacy-first internal automation stack¶

Use this when local control and internal data handling matter more than frontier-model convenience.

Shared internal knowledge stack¶

Use this when the company needs an internal assistant surface that people can actually use daily.

Multi-model routing stack¶

Use this when many models, providers, or routing policies are becoming a cost and reliability problem.

Finance and market intelligence stack¶

Use this when finance, market context, or target-account intelligence needs to become systematic.

What I would use first¶

If I were setting up an AI-driven company from scratch, I would start with:

Then I would add:

Browser Use when APIs are missing
mem0 when continuity actually matters
DeerFlow when research becomes a core workflow
LocalAI and llmfit when local/private inference is justified
AnythingLLM when teams need a shared internal AI workspace
LiteLLM or OpenRouter when model routing becomes operationally important
OpenBB when finance and market intelligence become recurring workflows

Sources / References¶

Contribution Metadata¶

Last reviewed: 2026-06-07
Confidence: high