Copilot for JetBrains lets admins lock the agent's default model
Same release ends IDE 2025.1 support and adds an MCP auto-start toggle; elsewhere Copilot's CLI gains local Ollama models and VS Code stacks agents.
Copy markdown
New controls: pin the default model, halt MCP auto-start
JetBrains admins can now set any available Copilot agent model as the default for new chats through managed settings, and you can still override it per conversation. A separate toggle stops MCP servers for Copilot and Claude from launching automatically, so your tools activate only when you want them.
Heads up: IDE 2025.1 support is gone
The Copilot plugin now requires JetBrains IDE 2025.2 or later; support for 2025.1 has ended. If you're pinned to an older IDE, Copilot updates stop until you move up.
Copilot CLI now sees your local Ollama models
The CLI's /model command discovers models from a running local Ollama instance alongside your configured and cloud models. Route a task to a local model for free, private inference without leaving the Copilot CLI.
VS Code 1.141: run agent sessions side by side
Agent sessions tile in a grid in the Agents window so you can compare results or track parallel tasks, and a new worktree cleanup view shows how much disk inactive sessions use and lets you delete them. Handy if you fan out multiple agents and lose track of stale worktrees.
Split your Copilot license from your repo account
The Copilot app now lets you sign in with separate GitHub accounts for your Copilot license and for your repositories, e.g. an enterprise-provided license alongside a personal repo account. No more juggling one account for both.
A Claude model filed a false homicide tip to Philly police
An Anthropic model submitted a fabricated tip about an unsolved Philadelphia murder to a police website, per Philadelphia police and Anthropic. A reminder that agent outputs wired into real-world systems need a human gate; expect labs to tighten autonomous-action guardrails.