# Raintree Architecture — How the Parts Fit Together

A plain-English explanation of the company workspace, configured connections, approved sources, and managed updates. See https://raintree.tech/architecture for the illustrated page.

## One request, end to end

A team message follows its assigned company setup, uses only approved business systems, pauses for approval when required, and returns the checked result to the team. Each role's setup defines the instructions, business systems, and approval rules Raintree must follow. Under the hood, each role runs as a governed agent loop: instructions and allowed systems load at the start, every system use is a structured, logged call, and any step that could change a company system pauses for explicit approval before it runs. Full flow at https://raintree.tech/architecture.

## Everyday use stays separate from admin control

Team members ask Raintree for work in the channels they already use and do not need access to the main controls. Company admins choose the roles, instructions, connected systems, and approvals. Under the hood, the workspace gateway creates isolated admin, company, and cron profiles. A resolved control-plane membership role gates dashboard routes and profile tool mounts; skill toggles and memory or skill write tools remain admin-only.

## See what works, then improve it

A clear work record saves the question, information used, action, approval, and outcome. Managers turn expectations into examples of good work that Raintree can check against. The proposed change is tried on fresh examples not used to create it, and a manager reviews the results before release. Under the hood, traces preserve model turns, tool calls, results, approvals, and outcomes; eval cases exercise the changed contract and only an admin-approved release moves updated policy into use.

## The agent remembers what matters

Raintree saves useful facts, decisions, preferences, and background from completed work, then brings back what matters for a later task. When a way of working proves useful more than once, it can become a repeatable company process. Under the hood, Raintree keeps saved context in a private memory store on company infrastructure. All of this changes the context and instructions Raintree works from, never the AI model itself. Learn more at https://raintree.tech/architecture.