How the agent works

The autonomous loop

Turn on «Autonomous mode» in the chat header and describe a task. The agent:

  1. Plans — analyzes the request and decides which tools are needed.
  2. Acts — calls tools: reads files, edits code, runs commands.
  3. Verifies — inspects each step's result and decides what's next.
  4. Repeats — until the task is done (with an iteration limit to prevent loops).
  5. Reports — sends a summary: what was done, which files changed, which commands ran.

Action feed

While working, every step is visible: tool name, arguments, result, duration. Long operations can be expanded for details.

Limits & safeguards

  • Loop detection — if the agent repeats the same actions, the loop is interrupted.
  • Iteration limit — upon reaching it, the agent is forced to produce a final report.
  • Three-tier patch application — exact match → line-based → whitespace-insensitive: edits apply reliably even to slightly changed files.
← Docs