Image: GitHub / github.blog GitHub Copilot for JetBrains Gets Session Memory, Ollama Support, and Enterprise MCP Controls
Two updates this month bring persistent memory across agent chats, local model support via Ollama, and centralized admin controls for plugins, MCP servers, and permission modes in JetBrains IDEs.
GitHub shipped two updates to Copilot for JetBrains this month that are worth covering together: an August 11 release adding persistent memory and Ollama as a BYOK provider, and an August 18 release giving enterprise administrators centralized control over plugins, MCP servers, and agent permission modes.
Session memory
Copilot can now retain and recall information across agent chat sessions in JetBrains. Previously, every new conversation started fresh. With memory on, Copilot holds project-specific context between sessions, such as coding conventions, preferred patterns, or ongoing task context, without requiring the user to re-explain it each time.
Memory is toggled on or off through Copilot settings. It works across agent chat sessions, not just inline completions.
Ollama as a local model provider
Ollama is now available as a BYOK provider inside JetBrains. Developers who run models locally can configure an Ollama endpoint and select models directly from within the IDE. Provider configuration and model selection are handled through the same model picker used for cloud-hosted Copilot models.
This gives teams a way to use locally-hosted models for code completion and chat without routing requests through external services, which matters for environments with strict data handling requirements.
Enterprise managed settings (August 18)
The August 18 update is focused on enterprise administration. Organizations on a Copilot enterprise plan can now apply consistent settings across all JetBrains users from a central policy configuration, with managed values taking precedence over developer preferences.
Plugin governance. Administrators can require specific plugins to be enabled or disabled via enabledPlugins. They can extend the approved plugin marketplace list with extraKnownMarketplaces or lock it down completely with strictKnownMarketplaces, limiting installations to sanctioned sources.
MCP server allowlists. The allowedMcpServers and deniedMcpServers settings give IT teams control over which Model Context Protocol servers developers can connect to from JetBrains. Organizations that maintain a list of approved MCP integrations can enforce it without relying on individual developer compliance.
Managed OpenTelemetry. Administrators can centrally configure the telemetry collector endpoint, protocol, service name, resource attributes, and content-capture policy. Admin values override what individual developers have set, ensuring telemetry routes to the right organizational collectors consistently.
Permission mode lockdown. Setting permissions.disableBypassPermissionsMode prevents the Copilot agent from using Bypass Approvals or Autopilot mode. For teams that require human approval on agent actions, this enforces that policy at the configuration level.
Other changes from August 11
The same update that added memory and Ollama also expanded Codex workflows: sessions now appear in debug logs, and Copilot CLI can be installed automatically from integrated terminals on macOS, Linux, and Windows. The model picker got usability improvements, file and folder references in chat were restored, and customization controls were repositioned for easier access.
August 11 changelog: github.blog/changelog/2026-08-11-copilot-memory-and-ollama-in-github-copilot-for-jetbrains
August 18 changelog: github.blog/changelog/2026-08-18-enterprise-managed-settings-in-github-copilot-for-jetbrains