Deerlight
Read and write projects, documents, skills and assets, with every change versioned
Send this to your AI
Read https://deerlight.lumenlan.com/skill.md and follow it to connect Deerlight
A product ecosystem for one-person companies
Inbox, planning, records, engineering, QA, keys — each Lumenlan product does one job, hands off cleanly to the next, and your AI can pick up the work directly. One Lumenlan account for all of it.
The crew
Each product does one job and does it fully. They share one Lumenlan account and hand off to each other by agreement.
Every mailbox, one workspace
Mail from all your accounts lands in one place, sorted by client, project and billing into the things that need you — read, search, reply and follow up without switching.
Knows where every project stands
Keeps each project’s goal, milestones and next step for you — who is waiting on whom and whether a step is really done, without relying on memory.
The current facts you and your AI share
Requirements, designs, conventions and decisions in one place — only the version that holds today. You organise on the web; your AI reads and writes through its connection, and every change keeps a version.
Every coding agent on every machine, in one place
Claude Code, Codex and OpenCode on your machines at home, at work or in the cloud — native terminals in the browser and on your phone. Work you hand off follows a plan, and it only calls you when it needs you.
An independent check before anything ships
Turns acceptance criteria into what must be checked, takes in the evidence, and returns a verdict the doer cannot rewrite.
Keys kept safe, still usable by your AI
Accounts, passwords and API keys in a vault you will actually use: one-click fill in the browser, and once you approve, your AI can use a key without ever seeing it.
Grow scattered ideas into finished work
Catch an idea when it arrives, then shape it step by step into a song, a letter or a story.
The moments that need you, on your phone
Reminders, approvals and sign-offs that need you personally, handled from your phone. Kite is available in Zhaoye today: fleets, terminals and alerts.
Done the same way, every time
Judgement goes to your agents; repeatable work goes here. It runs only on your machine and returns the same structured, checkable result every time.
How they work together
Each step belongs to one product and none overrules another: Robin owns progress, Deerlight owns the facts, Kite owns execution, Flicker owns the verdict.
Beemail
Turns incoming mail into things that need you
Robin
Sets goals and milestones, decides whether a step is done
Deerlight
Work starts from the requirements and rules that hold today
Kite
Agents on your machines get the work done
Flicker
Checks the evidence independently and returns a verdict
Keytail looks after your accounts and keys: your AI can use one only after you approve, and never sees it in plain text.
When a decision needs you personally, Zhaoye reaches you on your phone.
Hand it to your AI
Whatever you can do on the web, it can do once connected — same data, same permissions. Anything that costs money, goes out to others or cannot be undone, it explains first and waits for your yes.
Send this to your AI
Read https://www.lumenlan.com/skill.md and follow it to connect to Lumenlan
Your AI reads the guide and asks to connect
It gives you a link and an 8-character pairing code.
You check the code and approve
Approve only if the codes match — and grant only part if you like.
The credential goes straight to your AI
You never handle a key. Revoke it any time from the product.
Read and write projects, documents, skills and assets, with every change versioned
Send this to your AI
Read https://deerlight.lumenlan.com/skill.md and follow it to connect Deerlight
See machines and sessions, use memory, run remote work once you approve, and ask for your review
Send this to your AI
Read https://kite.lumenlan.com/skill.md and follow it to connect Kite
How we build
One job, one product
Each product owns one thing end to end. When it needs another capability, it hands off to the product that owns it.
You decide, AI does the work
Anything that costs money, goes out to others or cannot be undone is explained first and waits for your yes.
We only say what works
What you read in the product is what works today. When something fails, it tells you plainly instead of pretending.
Agents are users too
Whatever you can do in the interface, your connected AI can do too — same data, same permissions.
Every product stands on its own. Use two, and they can hand work to each other.