Microsoft Intelligent Terminal 0.2: Run Any AI Model Without Copilot

Microsoft shipped Intelligent Terminal 0.2 on August 10, adding local model support that lets developers run AI agents in Windows Terminal against any Ollama instance or OpenAI-compatible server without a GitHub Copilot subscription or cloud dependency. The update introduces Bring Your Own Model (BYOM), per-pane agents via the /agent command, WSL-aware execution, and OpenCode as a first-class option, positioning it as a vendor-neutral alternative to Warp Terminal's cloud-first AI backend.

Microsoft’s Intelligent Terminal 0.2 shipped on August 10 with local model support, and it changes the calculus on every other AI terminal out there. You can now run AI agents inside Windows Terminal against any Ollama instance or OpenAI-compatible server — no GitHub Copilot subscription, no cloud, no usage meter ticking. Point it at a local model, and your code stays on your machine. What Held Back 0.1 — and What 0.2 Fixes When Microsoft forked Windows Terminal at Build 2026 to create Intelligent Terminal https://devblogs.microsoft.com/commandline/intelligent-terminal-0-2-is-here-with-local-model-support/ , the reception was warm but conditional. The 0.1 release baked in GitHub Copilot as the default — and effectively only — agent, which meant you needed a Copilot subscription to get anything real out of it. For developers already paying for Warp or running local models through the CLI, it was a demo, not a tool. Version 0.2 addresses that directly. The Agent Pane now supports BYOM — Bring Your Own Model — which lets you configure any local or remote endpoint. The rest of the update builds on top of that foundation: per-pane agents, WSL-aware execution, and OpenCode as a first-class option. Setting Up Local Model Support If you have Ollama installed, setup takes about two minutes. Install the terminal, point the BYOM provider at your local endpoint, and you’re running agents against whatever model you have pulled locally. Install Intelligent Terminal winget install Microsoft.IntelligentTerminal Install Ollama if not already running winget install Ollama.Ollama ollama pull llama3.2 Then in the Agent Pane: open settings, go to the BYOM providers section, click Add New, set Base URL to http://localhost:11434 , and pick a Model ID. Any OpenAI-compatible server https://ollama.com works the same way — LM Studio, vLLM, or a remote endpoint you control. Once configured, the agent runs entirely on your machine. Per-Pane Agents and the Multi-Agent Workflow The other headline feature is per-pane agent selection via the /agent slash command. Each tab can run a different agent with its own isolated conversation history. In practice: Codex handling code generation in one tab, a local Llama model reviewing the same code in another, and OpenCode running a longer planning session in a third — all in the same terminal window. Built-in first-class options now include GitHub Copilot, Codex, and OpenCode. OpenCode is the interesting addition: it’s open source, supports local models natively, and with Intelligent Terminal’s BYOM layer underneath, you get a fully open stack with no subscription anywhere in the chain. WSL Integration: The Fix Windows Devs Actually Needed The 0.1 release ran agents in the Windows context, which created persistent path resolution issues for anyone working primarily in WSL2. Agents would suggest Windows-style paths inside Linux distributions, and filesystem operations would fail silently. 0.2 fixes this properly. Each profile in Intelligent Terminal can now be configured with its own agent — either Windows-hosted or a distro-local WSL agent. When you open a WSL profile, the agent executes inside that Linux distribution with the correct paths, working directory, and native tools. It’s the kind of fix that feels obvious in retrospect but matters enormously in daily use. What This Means for Warp Warp Terminal has been the default answer for developers who want AI built into their terminal. Its Agentic Development Environment, MCP integrations, and team workflows are genuinely good — but the Pro plan runs $12 per month, and Warp’s AI backend is cloud-first by design. Intelligent Terminal is still experimental — it’s 0.2, not 1.0 — but the architecture is vendor-neutral in a way Warp’s isn’t. Microsoft built it on the Agent Communication Protocol ACP standard under the Linux Foundation, meaning it’s not locked to any single AI provider. XDA confirmed it running with a local LLM in five minutes https://www.xda-developers.com/microsoft-new-intelligent-terminal-not-locked-running-local-llm/ out of the box. Microsoft’s bet is that the terminal becomes the primary AI agent interface for developers, and that it should be open. If they execute, Intelligent Terminal 0.2 will look like the release that made that bet credible. Source is on GitHub under MIT license https://github.com/microsoft/intelligent-terminal — install it, break it, and file issues.