GIDE 1.106.0: a rebuilt agent, Context Memory and a 21-language refactor engine

GIDE 1.106.0 is a rebuild of the agent at the heart of GIDE. Requests now flow through a plan-gate architecture, the agent checks and corrects its own work, and it can remember context well beyond a single session. Alongside that come a refactor engine for 21 languages, a much larger tool library and a redesigned code indexer.

At a glance

• A rebuilt agent core. Plan management, gate enforcement and a playbook engine.
• Reasoning, safety and self-correction. The agent judges its own edits and verifies them.
• Context Memory. Persistent memory that reaches beyond the in-session window.
• Refactor engine. 9 core operations across 21 languages.
• 30+ new agent tools. From LSP integration to notebook editing.
• Intelligent code indexer. Semantic, structural and symbol-level indexing.

A rebuilt agent core

What changed

Request handling was rebuilt around a plan-gate architecture: plan management, gate enforcement, and a playbook engine. Four built-in playbooks cover common code issues, such as missing packages, unused imports, incorrect import names and wrong usage patterns.

Reasoning, safety and self-correction

A new layer helps the agent notice when it’s going in circles, judge its own edits, classify errors, and validate fixes. Edits are verified after they’re applied, with automated testing, folder-scoped access and a dependency guard.

A new task planning model and an expanded model launcher improve how the agent coordinates and carries out its work.

Context Memory

Context Memory gives the agent persistent memory: it can store and recall what it has learned, cache file-level summaries, and archive conversations. That extends what the AI can draw on well beyond the in-session context window.

Refactor engine for 21 languages

A unified refactoring system supports 21 languages through 9 core operations:

• insert
• change value
• inline
• modify import
• wrap call
• add parameter
• rename symbol
• move to file
• extract function

The engine includes a dedicated scope analysis module. Native parsing support was also added for Bash, C, Dart, Dockerfile, HCL, HTML, Kotlin, PHP, R, Ruby, Swift and YAML.

More tools for the agent

More than 30 new tools cover user interaction, configuration, diagnostics, file operations, LSP integration, task management, context measurement, notebook editing, code search, web access and structured output. File-write capabilities were expanded too.

Intelligent code indexer

Code indexing was redesigned with semantic retrieval, structural fingerprinting and symbol-level indexing, for faster and more accurate understanding of your codebase. Older experimental retrieval modules were retired.

A better system prompt

A new system prompt gives more precise, context-aware answers across coding tasks. It replaces the older prompt templates.

Security and diagnostics

• Secure local authentication. Credentials are kept in a file-permission-protected local store, and the client and server verify each other with a protocol handshake before talking.
• Event logging. A structured event log with automatic rotation helps with diagnostics.
• Web sign-in. Signing in through the web companion now correctly establishes the desktop session.

Editor improvements and fixes

The editor layer gained support for agentic workflows, context gathering, edit validation, memory, permissions, caching and retry logic, plus an opt-in bug report bundler. Stability fixes landed across AI chat, mention completions, error quick-fixes, document handling, the chat view and notebooks.

For the short version of every change, see the GIDE changelog.