A coding agent is an AI system that writes, runs and fixes code toward a goal, using tools such as a file system, a terminal and tests, rather than only suggesting text. It plans steps, acts, checks the results and tries again.
For example, asked to add a column to a pipeline, a coding agent might read the pipeline's files, find the table definition, edit the SQL, run the project's tests, read the failure when one test breaks, fix the join it got wrong and run the tests again until they pass. The difference from a code completion tool is the loop: the agent acts, observes the result and decides what to do next, within limits set by the person using it.
Coding agents work through tools: reading and writing files, running commands, searching code and calling other services. Each step's result goes back to the AI model as part of the next prompt, so what the agent sees, such as file contents, error messages or sample rows, is sent to the model provider. Careful agents ask before actions that are hard to undo, such as deploying or deleting data, and keep a record of what they did so a person can review it.
General coding agents know a lot about code but little about a specific data platform: its merge syntax, its deployment unit or its permission traps. That gap is where many data pipeline mistakes come from, and it is the gap Emolysis addresses with platform-specific templates and tests. See also template slot, bring your own key and the security page.
Limits matter as much as skills: a coding agent should work in a sandbox or a development copy, never with production credentials it does not need, and a person should approve anything that deploys. See also golden test and MCP.
Emolysis is a coding agent specialized for data engineering and BI: it starts from tested templates and a platform profile, and proves its work with golden tests before deploying. See How Emolysis works. Other coding agents can use Emolysis through MCP.