Skip to main content

Explyt glossary

Agent Changes​

A panel with the files the agent changed in the current chat. You can review the diff there, accept the edits you want and reject the rest. For details, see Agent Changes.

Agent mode​

The role of the current agent and its baseline set of tools: General, Code, Ask, Test, Plan, Review, Debug or a custom mode. Do not confuse it with the connection mode: Personal, Community or Enterprise.

AGENTS.md​

A Markdown file with permanent project context and conventions. It guides the agent but is not a technical access restriction. For details, see AGENTS.md.

Auto Review​

An independent check of the task and the changes by a separate Review agent. The result is returned to the main agent in the current chat. For details, see Auto Review.

Chat compression​

Shortening the history of a long chat to free up space in the model's context. This is not Memory Bank: compression applies to the current chat, while memory covers knowledge across chats. For details, see chat compression.

Community, Personal and Enterprise​

The modes Explyt uses to connect to models. Personal uses the Explyt service and balance, Community uses your own API key, and Enterprise uses a corporate platform key. For details, see connection modes.

Edit Scope​

A restriction on the files and directories the agent can change in the current task. Edit Scope does not affect file reading. For details, see Edit Scope.

General​

The universal agent mode. It handles simple tasks itself and can pass complex ones to specialized subagents. For details, see agent modes.

MCP​

Model Context Protocol — a way to connect external tools and data to the agent: a browser, an issue tracker, a knowledge base or another service. For details, see MCP.

Memory Bank​

Project memory that persists across chats. The agent keeps entries in .explyt/memory/ and opens the relevant ones during later tasks. For details, see Memory Bank.

Model Routing Profile​

A named model profile for a subagent. A profile lets you use different models for fast, ordinary and complex subtasks. For details, see model profiles.

Orchestrator​

A mode for coordinating a large piece of work across specialized subagents. Orchestrator itself does not edit files and does not run ordinary tools directly. For details, see agent modes.

Review Results​

A window with the problems found during a review. Findings are tied to files and lines, and you can request a fix for them. For details, see review results.

Rule​

A Markdown instruction for the agent. A rule can be global or project-level and applies to files matched by filePattern. For details, see Rules.

Run Configuration​

A way, configured in a JetBrains IDE, to run an application, a test, a build or another task with the right SDK, environment and working directory. For details, see run configurations.

Skill​

An instruction for a repeatable specialized task. The agent can pick a Skill automatically, or the user invokes it with /skill-name. For details, see Skills.

Subagent​

A specialized agent that General or Orchestrator hands a separate subtask to: research, implementation, testing, review, planning or debugging.

Tool​

A technical operation available to the agent: reading a file, editing, searching for usages, running a test, working with the debugger or calling an external service. For details, see tool catalog.

Tools​

The menu for managing the tools of the current chat. In it you can allow a tool without confirmation, require confirmation, or forbid the call. For details, see Tools menu.

.readignore and .writeignore​

Files with gitignore-like path rules. .readignore forbids reading, and .writeignore forbids writing and creating files. To close a path completely, add it to both files. Support for the entire .gitignore syntax is not guaranteed. For details, see ignore files.