Weekly signal

This week (June 15–23, 2026) coding-focused agent tooling and runtimes hardened around two practical shifts: platform consolidation by major cloud vendors and rapid productization of agent-first developer surfaces. The biggest signals for engineering teams were Google’s forced migration from the community Gemini CLI to its new, closed Antigravity CLI (effective June 18), and a cluster of GitHub Copilot agent releases (desktop app GA, Agent Finder, AGENTS.md support, BYOK for Copilot CLI) that change how teams run and govern coding agents. OpenAI’s workspace/agent controls and rollout details also continued to shape enterprise deployment timelines.

What changed

1. Google retired Gemini CLI for most consumers and pushed Antigravity CLI as the new agent-first terminal and desktop platform; Gemini CLI and Gemini Code Assist stopped serving requests for free/Pro/Ultra individual users on June 18, 2026. Enterprises with paid Gemini/Enterprise access retain service, but consumer workflows must migrate to Antigravity (and read its docs for breaking differences).

2. GitHub shipped several agentic developer features mid‑June: the GitHub Copilot desktop app moved to general availability (June 17), Copilot added an Agent Finder surface, Copilot code-review added AGENTS.md support, Copilot CLI gained BYOK (bring‑your‑own‑key) model support, and per‑user AI credits metrics were added to usage APIs (June 17–19). These are practical changes for how teams run, meter, and secure coding agents.

3. OpenAI’s Workspace Agents and admin controls continued to be the baseline for enterprise agent deployments (general availability in Business/Enterprise/Edu earlier in May; free-preview and pricing timelines were reiterated into June). Admin-level safeguards, app permissions, and agent publishing controls are now available and should be part of rollouts.

What to do with it

1. If you or your CI/IDE workflows use Gemini CLI/Gemini Code Assist: plan an immediate migration (or test plan) to Antigravity CLI before or after June 18. Validate agent skills, extensions, and any scripts against Antigravity’s execution model; confirm rate limits and quota changes with your Google Cloud account.

2. For teams adopting GitHub Copilot agents: enable the Copilot desktop app in a staging workspace, test Agent Finder and AGENTS.md flows, and update billing/usage dashboards to consume the new per-user AI credits metrics. If you require enterprise model-control, evaluate BYOK for Copilot CLI now.

3. For enterprise rollouts: bake the agent admin controls (app permissions, action safeguards, RBAC) into your deployment checklist — OpenAI, GitHub, and Google now expect operators to own who can publish or schedule agents and what actions they can perform.

4. Update runbooks and threat models for always‑on or background agents: incidents from CLI transitions and surfaced telemetry show migrations and quota/billing gaps are the top operational hazards this week. Add validation tests that assert agent-scoped credentials are not leaked and that checkpoints abort long-running agent loops safely.

(See sources below for the primary product pages and changelogs cited.)