Who writes the scripts?
The agent. You describe in chat which data you need — it creates and maintains the scripts: writes new ones, fixes errors, adds sources. There is no code editor in the interface: all script work happens through a conversation with the agent, while the interface is there for running, scheduling and viewing results.
How is Data Agent different from manual data exports?
A manual export is one-off effort every time: someone has to write or fix code, remember to run it, watch for errors. Data Agent does all of it on its own: the scripts are already written, they run on demand or on a schedule, the history and logs of every run are kept, and the collected data is immediately ready for visualization in Analytics Agent.
How is Data Agent related to Analytics Agent?
Data Agent is the data supplier for Analytics Agent. Collection and preparation happen here: scripts, runs, the schedule, the store. Visualization happens in Analytics Agent: reports, charts and dashboards built on the databases Data Agent prepared. Together they form one pipeline: collect → prepare → present.
Analytics Agent gets read-only access to the data — databases can be viewed but not damaged. And when a report needs to be added, fixed or extended, Analytics Agent hands the work to Data Agent through its protected API (key access, every call verified): it puts or replaces a script, runs it, refreshes the data, and sets a schedule when needed. Only Data Agent changes the data — the store stays in order.
Where is the data stored?
In the Lakehouse — the /lakehouse folder on your server, organized by project and source. Each source folder holds: the raw Parquet files, a SQLite database for visualization, a README description and run logs. The data is yours: move it by simply copying the folder.
How is the schedule set up?
The schedule is one file, schedule.yml, in the scripts folder. The agent changes it at your request (“run every day at 8 a.m.”). Changes are picked up on the fly, without a restart, and runs missed to downtime never pile up — the schedule resumes cleanly, and a catch-up option recovers missed runs.
Do I need to know how to program?
No. Scripts, schedules and sources — the agent creates all of it from a verbal description. Python is only needed by the agent, and its environment is already built into the Docker image.
What do I need to run it?
A Docker host: one container, a scripts folder, a persistent data folder and a Lakehouse root. Chat with the agent is enabled with the address and API key of Xedant Agent.
Can I use it without Xedant Agent?
Yes: runs, the scheduler, the Lakehouse, data sources and browsing results all work on their own. The integration adds the agent chat — and the agent is exactly what writes and maintains the scripts, so without a connection to Xedant Agent there is no one to collect new data. Already collected data can be viewed and visualized in Analytics Agent.
How do I collect data from hundreds of pages?
Bulk collection is the job of harvesters: services for internet search, page download and text processing by models. You configure them once in the Harvesters section, and the agent launches them at your request — for example, “compare competitor prices across three hundred pages”. The result is visible in the Harvests section: how much is processed, how much was spent and how long is left. Whether paid keys are needed depends on the service: DuckDuckGo search works without a key, while search engines like Serper and AI models need a key and cost money. The cost is known in advance: every harvester states its price per query, per download or per token, and the statistics show what has been spent. Details in the Harvesters article.