Develop with the Createrun Agent or your own AI client.

Use the Createrun Agent in the Editor, or connect a compatible external AI client through MCP for supported application tasks. Review the resulting project changes before release. Sites and templates also have their own visual and file-based authoring paths.

Start for free See how a change is approved

Keep the development paths clear

External AI through MCP

Connect a compatible client to supported application operations. Inspect and review the resulting project changes. The client controls its model connection and interaction flow; Site and Template files have their own authoring paths.

The Agent in the Editor

Work inside the application workspace with the Createrun Agent. Review proposed changes in the context of the application.

Manual design and project files

Configure forms and workflows in the visual Editor. Work with site and template files for presentation, and write custom extensions where needed. AI use is optional.

FROM CONVERSATION TO RUNNING APP

Say what you need and watch it take shape.

The Createrun Agent provides an AI working environment inside the Editor. An external MCP client is another way to call supported Createrun operations using that client's model connection. These are distinct development paths. Describe an application change, inspect the proposed result, and verify it against your requirements. Supported operations and approval behavior vary by path and release.

COVERAGE

What the Agent can build for you today

Check the supported operations for your chosen AI path and installation. Permissions and review controls need to match the work you authorize.

Build

Forms & data models

The Agent creates screens, fields, and the tables behind them, and changes them later in the same conversation when the requirements move.

Design

Processes & approvals

Model multi-stage approvals visually and review each step before release. Workflows can preserve process state and resume after interruption when designed for that behavior.

Answer

Query & report

Ask a question of your data and turn the answer into a report your team can reuse.

Verify

Test safely

Changes are rehearsed in an isolated test mode with its own data, so production records stay untouched while you check the result.

Ship

Publish & roll back

Releases go out when you approve them. Changes are staged before they apply, with best-effort rollback.

Govern

Administer & audit

You decide who may do what, and you can read a complete record of who changed which object, and when.

Your engineers will find the tool reference on the Developers page.

REVIEW AND RELEASE

Review proposed changes before applying them.

The Createrun Agent supports a proposal, review, and application flow. Confirm the behavior of the operations you use, especially when connecting an external client. Your release process should include validation and an authorized production decision.

Related changes arrive as one reviewable batch.

Batching makes supported changes easier to review together. Inspect the affected objects and diagnostics, then confirm the result in the target environment. Recovery is best-effort and needs to be checked for the operation involved.

A proposed change is tested and reviewed before it is applied, with records and recovery options available for supported operations.

WHAT PROTECTS YOU WHEN SOMETHING GOES WRONG

Review, rehearse, and recover.

Staged changes. Use the proposal and review path for supported changes, and confirm when each operation takes effect.
Recovery options. Supported objects have recycle-bin or rollback paths. Check recovery coverage before applying a destructive change.
An isolated test mode. The Agent rehearses in a separate test environment, and your production data stays out of the rehearsal.
Review the action history. Inspect the records available for the operation and retain the evidence your release process needs.

YOUR MODELS, YOUR RULES

Choose the model path that fits your policy.

An AI client, a model provider, and model hosting are separate choices. Configure a supported connection that meets your organization’s requirements. Running Createrun and project memory on your servers does not by itself keep AI request data inside your network: context sent to an external model follows that provider’s service and data terms. A disconnected setup needs a compatible local model path. Confirm the client and provider combination in a test environment.

An AI client sends application context through a configured connection to either a hosted provider or a compatible local model.

Watch the Agent build a real application.

Bring an application requirement and ask to work through it with us. Review the proposed changes with us, then inspect the application objects, authored files and compiled output. We can also discuss how project knowledge and task tracking support work across AI sessions.

Book a demo