Practical tips on getting the most out of Xedant Agent, answers to common questions, and steps for fixing typical problems when something does not work the way you expected.
How to ask the AI — a few tips
- Be specific — the more details you give, the better the model understands the task. Instead of “fix the button”, say: “the submit button in the contact form does not check the email address — add that check”. A good request already holds half of the answer.
- Split big tasks into small ones — instead of “build the whole app”, start with one feature at a time. The model works best on clear, focused tasks.
- Use skills — pick the right skill for the task in the skill selector. A “code review” skill gives a different result on the same code than the “development” skill: each skill carries its own instructions for the model.
- Iterate — if the first answer is not perfect, ask for changes instead of starting over. “Make the heading bigger and change its color to blue” is faster than describing the whole requirement again.
- Give context — mention the file names, functions, or code samples the model should follow. For example: “Update the formatDate function in src/utils/date.ts so it understands the European date format.”
Common questions
How do I connect my own project?
Set your own environment variables (the project path, the Git repository URL, the credentials) and restart Xedant Agent. For details, see the Getting Started section and the installation guides for Windows or Docker.
Can I use several AI models?
Yes. Create several models in Settings — one for each AI model and its API key — and switch between them with the model selector next to the message input.
What happens if the model makes a mistake?
The model sees build errors and can fix them itself when AutoFix is on. You can also simply tell it what went wrong and ask for a fix: the model reads the error text from the build panel and uses it as context for the repair.
Can I move skills between projects?
Skills live in the .claude/skills/ folder inside the project and are stored together with it. To move a skill to another project, copy the skills folder, or ask the model to create similar skills from the existing ones as templates. For details, see the Skills section.
How do I back up my data?
Project files are in your Git repository: commit and push regularly. Chat history and settings live in a local database: keep a copy of the database file, and for permanent use you can set up an external database.
How private is my data?
Your code is processed locally — in your own container or instance. Only the chat messages you send to the AI provider (for example, Anthropic or OpenAI) go to the cloud. Analytics data stays local unless you yourself set up sending it to ClickHouse for unified reporting.
Connection problems
The AI does not answer: check that your API key is valid and that your provider balance has funds. Look at the key in the model configuration. If the server reaches the internet through a proxy (for example, in a corporate network), set the HTTP_PROXY and HTTPS_PROXY variables to the proxy address, and list the addresses to bypass in NO_PROXY — the requests to AI providers will then go through the proxy.
Cannot push or pull Git changes: make sure your Git credentials (login, personal access token) are set in environment variables. For details, see the Git section.
The container does not start: check the Docker logs for startup errors. The usual causes are wrong environment variables, a busy port, or missing configuration files.
The port is already in use: another program is using the port Xedant Agent needs. Change the port in the configuration, or stop the conflicting program.
Speed tips
- Big projects — the agent can work with large code bases, but speed depends on your container’s resources. For very large projects, give the container more memory and CPU time.
- Long conversations — very long sessions spend more tokens and can slow down. Start a new chat when the previous context is no longer needed for the task at hand.
- Build speed — when the full development environment is not needed, use the compact Docker image: builds and startup are faster.
- Database — with a large number of chats, PostgreSQL performs better than SQLite over time. Consider the switch when the project is growing fast.
Where to get help
If the tips above did not help and the problem is still there:
- Telegram — write to
@maxedantfor personal support (fastest); - Email — send a message to
support@xedant.com; - FAQ — see the FAQ page, which collects answers to common questions.
We usually reply within 12 hours, weekends included. Describe the problem, attach the error text and the steps that lead to it — that way we solve it faster.