AI application builder

Describe the change,
review it, ship it

X-AIApp turns a plain-language request into a reviewed code change and a verified deploy. A coding agent on your own worker machine writes the branch, a person reviews and merges it, and X-AIApp deploys it over SSH with health checks and automatic rollback.

Plain-language tasksHuman review before mergeGLM · DeepSeek · Qwen · ClaudeHealth-checked deploysBuilt-in Git hostingSelf-hosted
Self-hostedEnglish & 简体中文
X-AIApp task run, code diff and a verified deploy
Highlights

What X-AIApp does

From a request to a running service, with a person in charge at every gate.

Tasks in plain language

Describe an application or a change. Ask a question instead and the answer appears on the task.

Agents on your own workers

A coding worker runs the agent on the repository and pushes a branch. Workers only call out, so they can sit behind NAT or VPN.

Checks that fix themselves

After coding, the project's check command runs. If it fails, the agent gets the output and fixes it before review.

Review, request changes, merge

Read the diff, send it back with an instruction, or merge. A failed check cannot be merged without a written reason.

Code, docs and wiki together

Hosted Git with a code browser, a Docs tab for requirements and decisions, and a wiki for every project.

Servers you can see

Register a server once. A fixed, read-only probe reports the OS, runtimes, databases, nginx sites, certificates and ports.

AI

AI where it helps, automation where it matters

The model writes code. It never touches production.

🧠

Pick the model per task

Use the task's model, then the project default, then the global default. Every run records its model, tokens and cost.

🚀

Deploys never use an LLM

Upload, build, switch, restart and wait for the health check. If activation or the check fails, the previous release comes back.

🔒

Only merged work reaches servers

Workers can push only their own task branch, and those branches can never auto-deploy. Every deploy follows a human merge.

Guard-rails by design

Your code, your servers, your keys

Workers see a project only while running one of its tasks. Server and model secrets are encrypted with AES-256-GCM, host keys are pinned on first connect, and model keys reach a worker only inside the job that needs them.

Ready to try X-AIApp?

Bring one repository and one server; we'll take the first task to a live deploy with you.