GitHub expands Agent HQ to host Anthropic’s Claude and OpenAI’s Codex inside developer workflows

GitHub has turned Copilot into a formal multi‑agent workspace by adding Anthropic’s Claude and OpenAI’s Codex as selectable agents across GitHub, GitHub Mobile and Visual Studio Code for higher‑tier Copilot subscribers. The capability is surfaced through Agent HQ, a product view that treats AI assistants as interchangeable tools developers assign to tasks rather than as isolated apps to switch between. Teams can pick which agent to run on a given issue or pull request, attach agent runs to those artifacts for traceability, and view side‑by‑side outputs to compare quality and cost. GitHub treats each invocation of a third‑party agent as a premium Copilot request, making model experimentation subject to predictable usage charges. The launch dovetails with a broader platform trend: Microsoft’s Copilot SDK and Model Context Protocol (MCP) work enable centralized, remote CLI instances and grounded tool integrations, which let organizations host agents on servers, register MCP endpoints and reduce hallucination risk by connecting agents to project data and tools. That compatibility means teams can both run agents inside the editor and orchestrate them from centralized services while still observing Copilot licensing constraints. Parallel product moves — from Xcode’s agent hooks and OpenAI’s Codex macOS client to other vendor SDKs — show an industry momentum toward agents that keep long‑running context, orchestrate tests and compose skills across local and cloud services. For engineering managers and platform teams, Agent HQ reduces context switching by keeping prompts, artifacts and outputs in the same toolchain, but it also elevates operational tasks: secure token management, authorization for MCP endpoints, audit logs, edit transcripts, and IP and licensing reviews for generated code. Exposing multiple models inside a single workflow creates a mediated marketplace where providers are directly comparable on output quality, latency and cost; that will likely accelerate procurement and governance work around observability, provenance and cost metrics such as inference cost per query and premium invocation volume. In the medium term, GitHub’s orchestration layer strengthens its role as a gatekeeper for embedded coding AI, nudging both model vendors and platform operators to compete on integration, reliability and measurable developer productivity rather than on isolated benchmark scores.