AI, hosted with your apps.

Private AI agents for your apps, and a chat that builds and deploys them, in the workspace that runs everything else.

A private AI agent in three steps

No servers, no model accounts, no prompt code in your app. Only your app holds its key.

Your app sends a message with its key to your AI agent; the agent adds its instructions, guidelines and data and sends it to the model it picked for the task; the answer comes back the same way, in the OpenAI format.

Your app

ai.chat.completions.create(
  { messages }
)

Any language, one HTTPS call

Your AI agent

  • Instructions: who it is, how it answers
  • Guidelines: the rules it keeps
  • Data: Pricing.pdf, product docs

Added to every message, on Light Cloud

The model

  • Gemini 4 and Gemini 3
  • Gemma 4
  • gpt-oss

Change it per task, no code change

  1. 01Create itChoose AI agent on the Create page, or ask the chat for one. It is ready at once.
  2. 02Tell it what it knowsWrite its instructions and rules, add a price list, a web page or a PDF, and try it in the console.
  3. 03Connect your appYour app sends messages to one address with its key, in the OpenAI format. That is all the code.

The data you add is stored encrypted, Light Cloud keeps no prompts or answers, and the model providers do not train on them.

How a build goes

  1. 01Describe itSay what you want in plain words. It asks only what it cannot work out.
  2. 02Watch it buildEach step shows as it happens: the repository, the commit, the deploy, the logs.
  3. 03Get the linkThe code is in your own GitHub, GitLab or Bitbucket repository, and every push redeploys it.

With every paid plan

  • The chat30 messages a person a day, and its own AI use is on us.
  • AI agentsOn 13 models, at the providers’ own prices plus 5%.
  • Scheduled runsAgents that run on their own, every 15 minutes to once a day.
  • No AI subscriptionAI comes out of your plan’s included usage, like everything else.

Questions

Describe your next app

The chat opens with your projects beside it.