AI Agents Built Around Your Knowledge

Turn your private knowledge, workflows, and expertise into domain-specific AI agents that can answer questions, review information, and support real-world decisions.

Built for trusted knowledge and guided AI workflows.

Explore the Platform See How It Works

Beyond General AI

General AI models know public information. i80agent helps build agents that understand your business knowledge, workflows, documents, decisions, and expertise.

Knowledge Base Retrieval Expert Workflow Orchestration LLM
The platform

A different approach to AI agents

Most AI systems rely mainly on general-purpose LLMs. But real-world domains need more than a smart chatbot.

Hotels, clinics, restaurants, product teams, service businesses, and personal knowledge systems all have their own rules, vocabulary, documents, workflows, and decision patterns. A useful agent needs to understand that world.

i80agent is a platform for building domain-specific AI agents grounded in private knowledge, while also supporting guided expert workflows powered by LLMs.

Capabilities

Two Capabilities, One Agent Platform

Not every AI agent works the same way. Some agents mainly retrieve trusted answers. Others perform guided workflows such as review, analysis, recommendation, or decision support. More advanced systems can coordinate both.

Knowledge

Trusted Knowledge Retrieval

For known questions, the agent retrieves grounded answers directly from your structured knowledge base. In some cases, answers can be returned instantly without calling an LLM.

  • Semantic search and trusted retrieval
  • Grounded in your private domain content
  • Fast, cost-efficient, and easy to validate
  • Optional LLM refinement for explanation or ambiguity

Example: “What time is check-out?”

Workflow

Expert Workflow Processing

For complex requests, the agent applies a structured expert workflow powered by an LLM. Instead of only retrieving stored answers, the system guides reasoning using domain-specific logic and reusable workflows.

  • Reusable workflows for review, analysis, and decision support
  • Guided LLM reasoning based on domain expertise
  • Structured outputs for judgment-heavy tasks
  • Designed for cases where the answer is not already stored

Example: “Review this report and identify risks.”

A domain-specific agent can retrieve trusted knowledge, apply expert workflows, or coordinate both.

How it works

From user request to trusted response

i80agent supports different agent patterns. Some requests go directly to trusted knowledge. Others need expert workflows. More advanced use cases can be routed through an orchestrator.

1

User Request

The user asks a question, submits a document, requests a review, or starts a domain-specific task.

2

Knowledge Retrieval

For known answers, the agent retrieves trusted information from the domain knowledge base.

3

Expert Workflow

For complex tasks, the agent applies a structured workflow and uses an LLM for guided reasoning.

4

Orchestration

For advanced systems, an orchestrator can coordinate retrieval, workflows, tools, and external systems.

i80agent workflow showing knowledge retrieval, expert workflows, orchestration, and trusted responses

User Request

Question, document, task, or workflow

Route

Knowledge retrieval, expert workflow, or both

Response

Answer, analysis, recommendation, or action

Examples

Built for private-domain use cases

i80agent is designed for domains where general AI is not enough because the most important knowledge lives inside your documents, processes, experiences, or business context.

Personal Knowledge

Cooking stories, memories, journals, creative processes, research notes, and personal archives.

Hospitality

Hotel guest assistants, restaurant knowledge systems, service operations, and internal SOP access.

Review Workflows

Document review, risk analysis, data review, workflow guidance, and operational support.

Business Operations

Internal support agents, HR onboarding, product documentation, and team knowledge systems.

Why not just ChatGPT?

General LLMs are powerful, but they do not automatically understand your business, your documents, your workflows, your operational rules, or your internal knowledge.

Why domain-specific?

Each domain has its own vocabulary, context, rules, answers, and judgment patterns. A useful agent should be built around that specific world.

Why build in stages?

A useful agent does not need a perfect knowledge base or workflow system on day one. It can start small, learn from real usage, and improve over time.

Development Journey

Project Progress

What started as a personal experiment has grown into a broader exploration of how domain-specific AI agents can be built, refined, and eventually made accessible to others.

I am building this in my spare time alongside a regular job, so progress comes in waves and moves more slowly than I would like. But life, cooking, travel and curiosity are also part of the journey.

October 2024

Built the first working knowledge agent based on my personal cooking stories.

  • Built manually with AI assistance
  • Tested locally as a prototype for exploring private-domain knowledge retrieval
  • Structured around memory, retrieval, and answer generation
  • Focused on capturing personal knowledge rather than generic recipes
December 2024

Extended the same architecture to a real-world hotel assistant use case.

  • Hotel Assistant Agent deployed at Courtyard Shanghai International Tourism and Resorts Zone
  • Validated that domain-specific agents can support production environments
February 2025

Rebuilt the Alex Cooking agent into the Spontaneous Cooking by Alex set-menu dinner experience.

  • Adapted personal cooking knowledge into a guest-facing dinner experience
  • Tested the agent locally during live dinner settings
  • Explored how personal memory and storytelling can support hospitality experiences
2025

Continued improving retrieval quality, answer structure, embeddings, and practical usability.

  • Refined hotel-specific knowledge based on customer questions and feedback
  • Improved UI flow for a clearer guest assistant experience
  • Iterated on answer quality, coverage, and real-world usefulness
January – March 2026

Paused active development to rethink the foundation as AI tools and capabilities evolved faster than expected.

  • What makes an agent reusable across domains?
  • How should workflows be represented?
  • How can AI-assisted development reduce engineering overhead?
Starting April 2026

Began rebuilding i80agent using ChatGPT, OpenAI Codex, Claude Chat, and Claude Code, with a stronger focus on Retrieval Agents, Workflow Agents, and Orchestrator Agents.

May 2026

Rebuilt the i80.com website using OpenAI Codex without writing code manually, and refined the platform concept around private-domain knowledge and guided workflows.

What Works Today

  • Structured knowledge retrieval
  • Query-focused embedding strategy
  • Domain-specific answer cards
  • Retrieval and LLM routing
  • Basic orchestration logic
  • Early expert workflow framework

What Is Still Manual

  • Agent setup and configuration
  • Knowledge base structuring and ingestion
  • Workflow creation
  • Publishing and deployment
  • Use-case tuning and iteration

Current Focus

The goal is to evolve i80agent into a generic platform for building domain-specific AI agents that can retrieve knowledge, apply expert workflows, and coordinate real tasks.

  • A simpler agent creation flow
  • Better knowledge management UI
  • Confidence-based routing and decision logic
  • Reusable workflow templates for review, analysis, and decision support
  • Multi-domain scalability
  • A path from answers and workflows to triggered actions, APIs, and integrations
The story

Why I’m building this

It started with a simple realization: LLMs are powerful, but they do not know your personal or business knowledge.

That led me from cooking stories and memory into a broader vision: AI agents grounded in real domain knowledge and expert workflows.

I believe future AI systems will not just generate text. They will understand specialized domains deeply enough to assist with real-world work, decisions, and experiences.

Read My Road to i80
Alex starting the i80agent journey
Reflections

Research, Notes & Experiments

A living collection of experiments, product thinking, architecture notes, and reflections on building domain-specific AI agents.

Knowledge Agents Need to Know and Think
May 6, 2026

A practical framework for deciding when a knowledge agent should look up facts, apply expert workflows, or combine both.

Three-Layer Thinking for Building Knowledge Agents
May 3, 2026

How surface, essence, and philosophy can improve LLM use and help design better knowledge agents.

Is RAG Dead?
May 2, 2026

A practical reflection on why the concept behind RAG still matters, even as naive RAG becomes insufficient for domain-specific AI agents.

Beyond Vibe Coding: From Simple Prompts to Spec-Driven Development
April 26, 2026

Why fast AI coding breaks as systems grow, and how spec-driven development helps AI-generated software scale.

My Last Line of Code
April 3, 2026

A personal reflection on writing less code, designing more intent, and moving toward structured natural language.

Selecting the Best Embedding Method for Limited Domain Knowledge
May 21, 2025

A comparison of chunk-based and query-focused embedding strategies for private-domain agents.

Enhancing Query Retrieval Precision Through Optimized Embedding Text Selection
Nov 16, 2024

Research notes on how embedding construction affects semantic search precision.

Optimizing Retrieval in Private Knowledge Agents
Paper in progress

A comparison of query-focused embeddings vs. traditional embedding strategies for domain-specific AI agents.

Setting Up a Python Environment for AI Agent Experiments
Early practical note

A setup note from the first phase of the i80agent journey, when I was getting the Python foundation in place.

Exploring Domain-Specific AI?

i80agent is still evolving. I’m interested in connecting with people exploring knowledge agents, workflow agents, retrieval systems, or real-world AI applications built around private knowledge and domain expertise.

I hope i80agent eventually evolves into a platform that allows different forms of private knowledge and expertise to become useful AI systems.

Whether you are experimenting with a personal knowledge systems, operational workflows, domain-specific AI, or unusual use cases, I’d love to exchange ideas and learn how others are approaching this space.

Email Alex