Platform Features
One governed platform to build, connect and orchestrate AI.
Every capability runs under the same profiles and permission sets, so your AI works within your access rules instead of around them.
01 · MCP access
Make your app reachable from the AI assistants your people use
Let people ask questions of your business applications from Claude, ChatGPT or any assistant that supports custom MCP connectors. There is no separate integration project and no copy of the data.
Each app site can expose an MCP server from a single setting. Read and write are separate switches, so you can offer query access without allowing changes. Users connect the site as a custom connector and sign in with OAuth, and every query runs as that user. Two people asking the same question get different answers, because their profiles and permission sets decide what the assistant can see.
How it connects
02 · Builder agents
Build apps by conversation with the UES.IO App Builder, Schema Builder and View Builder agents
Describe what you need and get an app, a data model or a screen you can inspect, not a black box.
Three builder agents work inside the platform. The Schema Builder creates and changes collections, fields and relationships. The View Builder generates and edits views, including wires, components, layouts and merge logic, using your component catalog so screens match your design system. The App Builder combines both: given a description, it scaffolds collections, sample data, navigation and views in one conversation.
How it builds
03 · Coding agents
Let coding agents build on UES.IO
Turn a written request into a working, deployed feature, with permissions defined as part of the build.
Coding agents such as Claude Code and Codex, working with the UESIO Connector for Agents and the uesio-dev skill, can create collections, fields, views and permissions from a single prompt. The agent deploys to a dev workspace and checks the result in preview. Publishing to a live site is a separate step: a bundle version, switched on when you decide.
How it ships
04 · Orchestration
Orchestrate agents and choose the model for each step
Match model capability and cost to the task, and keep a record of what each run did and what it cost.
An orchestrator sequences agents into a workflow, and each agent has its own model, instructions and tool access. For example, a lower-cost model triages the data and a stronger model reasons only over the flags it produces. Runs execute as the signed-in user, and checks outside that user’s access are skipped and reported. Every run shows its tool calls, tokens and cost by model.
Configure your own
Governance
One governance layer under every feature
Permissions, audit and deployment choice apply the same way whichever feature you use.
Read how UES.IO handles AI governance and trust and security.
Frequently asked questions
Can I allow queries without allowing changes?
Yes. MCP read and write are enabled separately.
Can the assistant see data the user can’t?
No. Requests run as the signed-in user, under their profile and permission sets.
Does the coding agent change my live application?
In the workflow shown, it deploys to a dev workspace. Publishing to a site is a separate step.
Are agent-built apps different from hand-built ones?
No. The agents produce the same declarative definitions, under the same schema rules, component contracts and deployment path.
Which AI assistants can connect?
Claude, ChatGPT and any other assistant that lets you add a custom MCP connection. Each user signs in with their own UES.IO account.
Which models can I use?
UES.IO supports 14 model providers, and each agent can use a different one.