Your front door for AI agents.
Lodestar gives a small service business an agent-readable presence so AI assistants can discover it, answer questions about it accurately, and send it leads.
Starter $19/mo, Pro $39/mo (adds the AI suite). Every plan starts with a 14-day free trial.
The problem
People increasingly hand the lookup to an AI agent instead of a search bar. If a business isn't machine-readable, the assistant answers from guesses or leaves it out of the answer entirely.
The approach
Lodestar keeps a synced set of agent-facing surfaces — public files plus a live MCP server — so any assistant can find the business, answer strictly from facts the owner controls, and pass real inquiries back to one console.
What's inside
Public agent surfaces (llms.txt, JSON-LD, discovery doc)
Per-business MCP server with read tools
Grounded ask tool that cites its source
AI-triaged inquiries inbox
Agent analytics dashboard
Agent-readiness score with prioritized gaps
AI content generation (summary, blurbs, FAQs)
Import your business from an existing website
03From the lab
The decisions, trade-offs, and engineering notes behind Lodestar — written as we build.
AI assistants now answer customers directly, from whatever they can read about your business — and you can't see the ones you lose. Here's the five-minute test to read the page your customers already do.
Building an AI tool to answer factual questions about a business meant teaching it to refuse. A grounded pattern worth stealing: cite the source field, design the "I don't know," and log the misses.
"Agent-readable presence" sounds like a black box. It's mostly a few plain files — an llms.txt, a schema.org JSON-LD block, a discovery doc. Here's what's actually in them for a small business, and how to write one yourself.
More from Eastbase
