GitHub Adds Persistent Memory and Ollama to Copilot for JetBrains
The IDE plugin can carry selected context across agent chats, connect to Ollama as a bring-your-own-key provider, expose Codex sessions in debug logs and accept tighter enterprise controls.
GitHub released an update to Copilot for JetBrains on August 11 that adds persistent memory across agent chats, Ollama as a bring-your-own-key model provider and new controls for enterprise administrators. The official changelog also expands Codex workflow visibility and simplifies Copilot CLI setup inside supported JetBrains environments.
The practical change is a wider range of choices about where coding-agent context lives, which models a developer can select and what an organization can govern centrally. Teams using IntelliJ-based IDEs can reduce repeated project briefing between conversations, while administrators gain settings for plugins, MCP servers, permission behaviour and telemetry.
Context can persist between agent chats
Copilot memory can now retain and recall useful information across agent chat sessions. GitHub presents the feature as a way to carry project details or developer preferences between conversations instead of asking users to supply the same context repeatedly. A toggle in the Copilot settings portal lets users manage whether the feature is active.
Persistent context can make a long-running coding workflow less repetitive, but it also changes the mental model of a chat session. A new conversation is no longer necessarily a clean slate when memory is enabled. Developers working across clients, repositories or security boundaries will need to understand what the feature stores and when remembered information is applied. The announcement confirms the capability and toggle, but does not provide a detailed retention model in the fetched evidence.
Ollama becomes a model-provider option
The JetBrains plugin can now use Ollama as a BYOK provider. Developers can configure the provider and select an Ollama-served model throughout the JetBrains experience, adding a local-model route alongside the other model options available through Copilot.
That is useful for teams that want more control over model choice or already operate Ollama in their development environment. It should not be read as a blanket privacy or offline guarantee: the changelog documents provider configuration and model selection, but does not enumerate supported Ollama models, networking requirements, feature parity or data-handling behaviour for every configuration. Those details remain checkpoints for the product documentation and real deployments.
Codex, CLI and enterprise controls expand
GitHub says Codex sessions now appear in agent debug logs. Codex workflows also gain updated permission modes and project customization through instructions and skills, giving developers more ways to shape agent behaviour and more visibility when troubleshooting a session.
The plugin can automatically install Copilot CLI from an integrated terminal on macOS, Linux and Windows. That reduces a setup step for developers who want terminal-based agent workflows alongside the IDE experience. The release also restores file and folder references in Copilot, Claude and Codex chat inputs and reports reliability work across MCP execution, approvals, terminal output, customizations, cloud agents and diff-based editing.
For managed organizations, server-based settings cover plugin availability, MCP server access, permission-bypass behaviour and OpenTelemetry configuration. These controls matter because a coding agent can reach beyond text completion into tools, terminals and external context. Central policy can make those capabilities more consistent across a company, although the announcement does not evaluate whether any particular policy configuration is sufficient for a team's security requirements.
What to watch next
The next useful evidence will be the supporting documentation for memory scope and deletion, Ollama compatibility, and the interaction between enterprise settings and developer-level choices. Reliability claims also need experience across real repositories and MCP configurations; the official changelog records fixes, not an independent test suite or failure-rate comparison.
The supported conclusion is therefore specific: GitHub has shipped persistent Copilot memory, Ollama provider support, expanded Codex workflow controls and related management changes in its JetBrains plugin. How much those features improve continuity, privacy or reliability will depend on configuration and subsequent evidence.
Status
Confirmed. Availability and feature scope come from GitHub's official changelog. Internal confidence is medium because the article relies on one official source and the operational benefits have not been independently measured in the fetched evidence.
Sources
Update note: Last reviewed 2026-08-12. We will revise this post when GitHub publishes material changes to memory controls, Ollama compatibility or enterprise policy behaviour.
Sources
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.