Where your teams use AI, on your own cloud. Agents, documents and workflows in English and Persian, with approvals, data rules and budgets.
Every Monday it reads last week’s figures from your database and compares them with targets and the week before. A manager then decides each proposed action.
Staff ask in Persian or English. The answer quotes the passage it used and links to it.
It answers customers or staff where they already talk. You decide how much each person may ask and what it may use.
Mark each document public, internal, confidential or restricted. Then choose which documents each model provider may receive.
A form or a webhook starts a workflow. It fills a table row, asks an approver and tells the requester on Telegram, Slack or email.
Each tool can be allowed or denied for a person or an agent. Agents can act under identities of their own, and one Actions view shows what each did and who approved it.
Local models run through Ollama, vLLM, LM Studio or LiteLLM. Each agent has a monthly budget, checked before every run. Runs through the Gateway cost what the Gateway charged.
Workflows read company databases through saved queries that may only read. Claude Code and Cursor can work in the Workspace through its own MCP server.
A draft stays private until it is published. Every published version can be compared and restored.
Rated answers and knowledge gaps show where an agent needs better documents.
People see only the agents, documents and tools their roles allow.
Search handles Persian letters, digits and half-spaces, so a word is found however it was typed.
Running code and a browser for agents need a separate server, so they stay off until one is added.
Tell us what you want it to do. We show it in a demo, then set it up with you on your own cloud, with the Gateway.