Using SolidX AI Agent
Build modules, models, and features through natural language with the SolidX AI agent.
Once your SolidX project is running, you can start the SolidX AI agent to build modules, models, and features through natural language.
Prerequisite: bootstrap a project first
This guide assumes you have already scaffolded and started a SolidX project using solidctl create-app and solidctl start:dev, as described in the main Getting Started guide. The agent operates on an existing project - it does not create one.
Python requirement
The SolidX AI agent requires Python 3.12+. Verify with python --version. Python is required because the SolidX agent runtime is Python-based. On first run, solidctl agent start auto-installs the agent's Python package into ~/.solidx/venv if solidx-agent is not already in PATH.
Start the agent server
solidctl agent startOpen the Chat UI at http://localhost:8768.
Run a single task (CLI)
solidctl agent run "Build a CRM with Lead and Contact models, add name/email/phone fields"Options:
| Flag | Default | Description |
|---|---|---|
-m, --mode | native | Tool mode: native or mcp |
-l, --log-level | INFO | Logging level |
Agent commands
| Command | What it does |
|---|---|
agent start | Start the AI agent server |
agent run "<task>" | Run a single agent task |
Full reference: solidctl Commands
Next Steps
- Explore the agent Chat UI at
http://localhost:8768. - Connect an external MCP client to SolidX: see Using SolidX MCP Server.
- Go deeper: read the Reference and Recipes.

