GitHub Copilot August 24 Update: Slack Sessions, Rust CLI, and VS Code 1.135

GitHub shipped its August 24 Copilot release on August 28, 2026, touching every surface of the product: Slack and Microsoft Teams get shared agent sessions, the CLI gets a native Rust runtime for speed, the Copilot app's Customize tab reaches general availability, and VS Code 1.135 brings per-model chat usage tracking. Visual Studio adds model pinning and org-wide custom agent sharing. Here is what changed and what it means for teams building with Copilot day to day.

What happened

Area	Key change
Slack and Teams	Shared agent sessions via @GitHub mention
Copilot app	Customize tab now generally available (MCP, plugins, skills, canvases)
Copilot CLI	Native Rust runtime for faster performance; session restore added
JetBrains	Enterprise controls for plugins, MCP, telemetry, and permission modes
VS Code 1.135	Per-turn model chat usage details; resume sessions from other apps
Visual Studio	Pin/hide models, compare costs, share custom agents org-wide

Copilot in Slack and Microsoft Teams

Teams can now mention @GitHub inside a Slack or Teams conversation to open a shared agent session. Everyone in the thread can follow along, ask questions, and steer the session. That brings Copilot’s agent capabilities into the chat tools most teams already live in, rather than requiring a context switch to a separate IDE.

Copilot app: Customize tab goes GA

The Customize tab, previously in preview, is now generally available. It consolidates MCP servers (Model Context Protocol, a standard for giving AI models access to external tools and data), plugins, skills, and canvases in a single place. You can also turn Azure DevOps issues and pull requests directly into Copilot sessions from this tab.

Other app updates include experimental Windows Subsystem for Linux (WSL) support, the ability to split and move tabs for side-by-side work, and a right-click option to send any browser preview to an external browser.

CLI gets a Rust runtime

The Copilot CLI now runs on a native Rust runtime, which GitHub says delivers significantly faster performance. The terminal interface stays in TypeScript. Two new config options, defaultMode and defaultPermissionMode, let you set your preferred execution and permission settings at the start of every session. The CLI can also restore sessions that ended uncleanly, including those interrupted mid-turn. Plugin, MCP server, and skill management gets updated experiences inside /plugin, /mcp, and /skills.

JetBrains: enterprise controls land

Enterprise administrators can now enforce consistent Copilot settings across JetBrains IDEs, covering plugins, MCP servers, telemetry, and agent permission modes. This closes a gap that existed for teams standardizing on IntelliJ, PyCharm, or other JetBrains tools.

VS Code 1.135

The VS Code 1.135 release adds a few things worth noting. You can continue a Copilot or Claude agent session that was started in a different application directly inside VS Code. A new “second opinion” feature routes a query to a complementary model to catch edge cases or details the first model missed. A redesigned single-pane Agents layout simplifies session controls. Most practically for budget-conscious teams: you can now see detailed chat usage broken down by model for every single chat turn.

If your team uses AI integration across multiple tools, the cross-app session continuity in VS Code 1.135 is one of the more useful workflow changes in this release.

Visual Studio additions

Visual Studio users can now pin preferred models and hide ones they do not use, with a side-by-side view that compares model capabilities and costs. Reasoning effort is now adjustable per task. Custom agents can be shared across an entire organization to standardize workflows. A plan consumption view lets you monitor premium model usage before hitting limits. The Git agent can review uncommitted changes or individual commits before a pull request is opened.

Why it matters

The Slack and Teams integration is the headline for team leads. Copilot sessions have historically been single-user by default, so moving them into shared channels is a meaningful shift. It also means decisions made during an AI session are visible to the whole team without anyone having to paste screenshots.

The Rust CLI runtime matters for anyone running Copilot in terminal-heavy workflows or CI pipelines where startup time accumulates. GitHub did not publish specific benchmark numbers, but “significantly faster” in a CLI context usually means measurable seconds, not milliseconds.

The per-model usage view in VS Code 1.135 is quietly important. As teams mix Claude, GPT-4o, and other models inside Copilot, knowing exactly which model generated which response and what it cost per turn is the kind of data you need to manage spend and audit outputs.

You can track further Copilot developments alongside other AI news on the Lumien news feed.

Our take

GitHub is clearly trying to make Copilot the connective tissue between where work happens (Slack, Teams, Azure DevOps) and where code gets written. That is a smart bet. Developers already bounce between chat and IDE dozens of times a day, and reducing that friction has real productivity value.

The Rust CLI migration is a good engineering call. TypeScript terminal interfaces on a Rust core is a common pattern now (Zed, Ruff, and others use it), and it tends to pay off in responsiveness. The session restore feature is the kind of quality-of-life fix that should have been there already, but better late than none.

The model cost comparison in Visual Studio is the feature we would highlight most for business clients. Once you can see model costs per chat turn, you can actually make informed decisions about which model to use for which task type, rather than just defaulting to the most capable one and absorbing the bill. Teams building on GitHub Copilot should check the web development services page if they want help structuring those workflows.

The one thing missing from this release: clearer documentation on what “shared agent sessions” in Slack actually allows. Can a non-developer teammate approve a change? Can the session push a commit? The announcement does not say.

What to do about it

1. Update the Copilot CLI and set defaultMode and defaultPermissionMode to match your team’s security policy before rolling it out to others.
2. Open the Copilot app’s Customize tab and audit which MCP servers and plugins are active now that the tab is GA.
3. In VS Code 1.135, enable the per-model usage view and run it for one sprint to get a baseline on which models your team actually uses and at what volume.
4. If your org uses JetBrains IDEs, push the new enterprise controls through your admin console to enforce consistent permission modes across the team.
5. Test the Slack or Teams @GitHub integration in a low-stakes channel before rolling it out to the whole engineering team.

Pin model costs in Visual Studio now. Waiting until you hit a premium limit is the expensive way to learn which models your team over-uses.

Source: GitHub Changelog