Included with The AI Automators membership

The Full Stack
AI Agent Platform

Your agents. Your knowledge. Your infrastructure.

A complete AI workspace with the source code: cited answers over your documents, research agents that deliver finished files, a living knowledge wiki, skills, tools and code execution. Run it as it is, or build your own product on top of it.

ReactPythonFastAPISupabasePostgresDockerOpenAIOllamaLM Studio
  • Full source codeReact, FastAPI and Supabase you own and change
  • Runs where you chooseFully local, self-hosted or in the cloud
  • Build and sell on itCommercial and client use included
  • Keeps getting betterNew features as we build them on YouTube

The workspace

A chat workspace that feels like the best AI apps

A calm, fast interface people already know how to use, with the controls power users expect.

  • Streamed answers keep going when you switch threads or pages mid-generation
  • Choose the model, thinking and effort for each conversation
  • Three modes from one composer: Ask, Deep research and Contract review
  • Paste or upload images straight into a message
  • Thread search, inline rename and automatic titles
  • A context window meter, and compaction that keeps long conversations within limits
Model, thinking and effort, per conversation.

Answers with evidence

Every claim, traced back to the exact sentence

Answers cite verbatim quotes from your documents. Click a citation and the source opens beside the chat with the passage highlighted.

  • Quote-based inline citations with hover previews and a source panel
  • Click-through highlighting in PDFs, rendered Markdown (tables included) and plain text
  • Web search citations snapshot the cited page, so the evidence never moves
  • Optional citation checking marks each claim as supported or partly supported
  • Copy an answer or re-check its citations from the actions under it

Show the work

See what the agent did, in plain language

Every search, tool call and sub-agent appears as a readable step, so users trust the answer and builders can debug it.

  • Steps named for people, not logs: Searched documents, Ran Python
  • The exact query behind each search and the documents it found
  • Open a step to see its request and response directly
  • Sub-agent panels and code runs persist across reloads

Agents that finish the job

Hand over a task. Get back finished files.

In Deep research mode the agent writes a plan, works through it with sub-agents and tools, and returns real deliverables you can open and download.

  • A live plan that ticks off each step as the agent works
  • Sub-agents with isolated context for focused research
  • A persistent workspace of files for every conversation
  • Python in a Docker sandbox for analysis, charts and file generation
  • Generated files preview inline and download in one click
The code it ran, one click away.

Agent harness

Multi-phase workflows you can adapt to any domain

The harness runs long, structured jobs phase by phase. The included Contract review shows the whole pattern: intake, classification, playbook, clause extraction, risk analysis, redlines and an executive summary.

  • Eight phases with visible progress and saved state
  • Delivers a Word report, redlines and a full risk analysis
  • Pulls your own playbook from the knowledge base
  • Swap the phases and prompts to build a harness for your own vertical

Knowledge base

Ingest anything. Find it instantly.

Drag in PDFs, Office files, HTML, Markdown and images. Docling turns them into clean, structured text for a complete retrieval pipeline.

  • PDF, DOCX, PPTX, XLSX, HTML, Markdown and images, with an OCR fallback
  • Structure-aware chunking and pgvector embeddings
  • Hybrid semantic and keyword search, with reranking
  • Content hashing skips duplicates; an LLM extracts structured metadata
  • Folders, bulk move and delete, staged progress and one-click retry
  • Agent tools to explore the library: ls, tree, grep, glob and read
Preview, text, details and chunks for every file.

Go beyond document chat

A living knowledge base, compiled from your documents

Choose your folders and the wiki builds linked, source-backed pages from them. When a source changes, the pages built from it update too.

  • Pages for concepts, entities and references, with footnoted claims
  • An editor with AI-proposed edits you review, plus full revision history
  • An Obsidian-style graph to explore how everything connects
  • Health checks flag duplicates and failed updates, with Fix with AI
  • Chat reads the wiki, and any good answer can be saved to it
  • Export the whole wiki as a folder Obsidian can open
Every claim footnoted to its source.

Skills

Teach the agent how your team works

Skills are reusable instructions the agent loads on demand when a chat needs them. Write one in the editor or import one in seconds.

  • SKILL.md instructions with their own files and linked document folders
  • Import from a URL, including skills.sh and GitHub
  • Keep skills private or share them with everyone
  • Try any skill in chat with one click

Under the hood

The engineering you would otherwise spend months on

The parts every serious agent platform needs, already wired together and ready to extend.

Tools and integrations

  • MCP client: connect external tool servers such as GitHub
  • A unified tool registry with on-demand discovery through tool_search
  • Code Mode: sandboxed code calls platform tools through an HTTP bridge
  • Web search when your documents do not have the answer
  • Parallel tool calls for faster research

Any model, your choice

  • OpenAI, OpenRouter or any OpenAI-compatible API
  • Local models through Ollama or LM Studio
  • Run calls on a ChatGPT or Codex subscription
  • Optional Jev router sends simple questions to a faster, cheaper model and shows which model answered
  • Jev steerable reranking, or any Cohere-compatible reranker

Privacy and control

  • Supabase sign-in, with row-level security in the database
  • Private libraries, plus folders owners can share with everyone
  • PII redaction with Microsoft Presidio: realistic surrogates, reversible anonymisation
  • Run fully locally with self-hosted Supabase and local chat and embedding models

Observability and evals

  • Langfuse or LangSmith, swappable by configuration
  • Sessions, users, cost and token usage for every run
  • Prompt management with an in-code fallback
  • User feedback scores and a review queue
  • Datasets and LLM-judge evals, with PII and secrets masked in traces

Designed as one product

Polished on every screen, ready for your brand

One design system across chat, documents, skills and the wiki. Dark by default, a light theme of equal quality, and a phone layout treated as first class.

  • Shared components for every dialog, sheet, menu and status
  • Responsive from a 390 px phone to a wide desktop
  • Accessible controls with visible focus and labelled icons
  • Swap the logo and brand colour to make it yours

The stack

Mainstream tools, no lock-in

Built on technology your team and your AI coding tools already know, so every part is easy to read, change and deploy.

Frontend
React, TypeScript, Vite, Tailwind, shadcn/ui
Backend
Python, FastAPI
Database
Supabase: Postgres, pgvector, Auth, Storage, Realtime
Documents
Docling
Tools
Model Context Protocol, Docker code sandbox
Privacy
Microsoft Presidio, spaCy, Faker
Models
OpenAI, OpenRouter, Ollama, LM Studio, any OpenAI-compatible API, ChatGPT and Codex subscriptions
Observability
Langfuse or LangSmith

The members license

Build on it. Sell what you build.

The platform is a foundation for your own solutions. The license is written to let you build a business on it.

You can

  • Use it for personal or commercial work
  • Modify the code however you like
  • Deploy it in production for yourself or your clients
  • Build products and services on it and keep all the revenue
  • Take on unlimited client projects and hand clients the source, license included

You cannot

  • Publish the source code publicly
  • Share access or copies with people who are not members
  • Resell or give it away as a template, boilerplate or starter kit
Yours to keep

After the 7-day refund period your license becomes perpetual. It stays with you even if your membership ends, for every version you accessed as a member.

A plain-English summary. The Members License Agreement included with the code is the version that applies. Not sure about something? Email info@theaiautomators.com.

Your way

Four ways to use it

  1. Use it as it is

    Clone it, add your branding and run a private AI workspace for your team. Pull new releases as they land.

  2. Fork and extend

    Start from any release and add the features your use case needs, with Claude Code or Codex doing the heavy lifting.

  3. Make it your product

    Take the foundation in your own direction: a vertical AI tool, a client solution or a SaaS. The planning docs help you port new features across.

  4. Build it from scratch

    Follow the AI Builder Series on YouTube with the PRDs for every episode and learn by building the whole system yourself.

Questions

Before you start

Is it production ready?

It is an evolving reference implementation and a learning foundation, actively developed and expanded. Many features are optional and switched on through configuration. Test it against your own use case and harden what you ship, as you would with any codebase you take on.

Do I need to be a developer?

It helps to be comfortable running a few commands. The platform was built with Claude Code across our AI Builder Series, and the repository includes context files for Claude Code and Codex plus a step-by-step install guide, so an AI coding tool can help you set it up and change it.

Can it run entirely on my own machine?

Yes. Use self-hosted Supabase with Docker and local models through Ollama or LM Studio for both chat and embeddings, and keep the optional external tools and tracing switched off.

Can I use it for client work and sell what I build?

Yes. You can deploy it for unlimited client projects, build products and services on it and keep all the revenue. What you cannot do is publish the source or resell it as a template or boilerplate.

What happens if my membership ends?

Once the 7-day refund period has passed, your license is perpetual. You keep the right to use every version you accessed as a member. Rejoin at any time to get the updates you missed.

How do I get access?

Join The AI Automators, then request access to the private GitHub repository from the community. You get the source, the setup walkthrough and every update while you are a member.

Start with a platform.
Build what comes next.

Get the Full Stack AI Agent Platform, every other starter app, our courses and live workshops with your membership.

Backed by our 7-day money-back guarantee.