Google Antigravity release notes: 2.1.4 adds PDF support and /btw

Google released Antigravity 2.1.4 on June 11, 2026. This is not a giant product launch, but it is a useful Google Antigravity release notes signal: Antigravity now has a clearer quota screen, PDF support in messages, a new /btw path for side questions, and several fixes that make agent work easier to review.

Google Antigravity is Google's agentic development environment: a workspace where coding agents can plan, write, run, and check work across a project, terminal, and browser. For Nordic teams testing agentic development, small changes like these matter. They often decide whether a tool feels like a demo or something you can put inside a real workflow.

Source: Google Antigravity Changelog — version 2.1.4, June 11, 2026. Google also says new versions roll out gradually and may take a few days to reach all users.

Source: Google Developers Blog: All the news from the Google I/O 2026 Developer keynote — describes Antigravity 2.0 as Google's agent-first development platform.

Google Antigravity release notes: what changed in 2.1.4

The most practical Antigravity 2.1.4 changes are four things:

• The quota screen in the Models tab of Settings now gives a clearer view of used credits versus remaining credits.
• /btw lets you ask a side question inside an ongoing conversation to an ephemeral agent that has context from the current conversation.
• PDFs can be attached to messages to Gemini models, either by dragging and dropping PDFs from the filesystem or by using the media option in the Add Context button in the input box.
• cmd/ctrl+F searches visible text in the conversation view.

There are workflow improvements too: breadcrumbs in the file viewer, nested subagents in the overview pane, and a cleaner project flow where you can name a project during creation and sort conversations by worktree.

Source: Google Antigravity Changelog — the "Quota Screen Redesign and PDF Support" section lists /btw, PDF attachments, cmd/ctrl+F, breadcrumbs, nested subagents, and project improvements for version 2.1.4.

Why this matters for Antigravity work

The quota screen may be the least flashy change, but it could be the most useful one for teams. If several people run Gemini models, subagents, and background jobs, someone needs to know how much capacity is left before starting the next larger refactor or document review.

PDF support makes Antigravity more relevant to ordinary business workflows, not just code. A requirements document, technical appendix, proposal, policy, or workshop note can travel with the agent task. The question then shifts from "can AI read the document?" to "did we give the agent the right scope, the right access, and a human reviewer for the result?"

/btw is useful for a different reason. A side question to an ephemeral agent can keep the main thread from drifting. That helps when you want to check an assumption, ask for a short explanation, or compare two options without changing direction for the whole task.

Checks before you let more agents in

Antigravity 2.1.4 includes two fixes that point in the right direction for control. MCP stability improves, and mcp_config.json now accepts url in addition to serverUrl. MCP, the Model Context Protocol, is a way to connect agent tools to external systems in a more standardized form. Google also adds .vscode and .cache to the list of sensitive paths that require explicit user confirmation before an agent can access them.

For Hammer readers, the practical read is simple: build the integration, but make it reviewable. Use environment variables or secret managers for secrets, scoped API keys, least privilege, output redaction, approval gates, and logs that show what the agent did.

Source: Google Antigravity Changelog — the 2.1.4 fixes mention improved MCP stability, mcp_config.json with url or serverUrl, and .vscode plus .cache as sensitive paths that require explicit confirmation.

Short example: use the new Antigravity feature

Human step: open Antigravity once version 2.1.4 has reached your account. Attach a PDF to the message by dragging it in from the filesystem or by using the media option in the Add Context button. If you need a side question during the work, type /btw in the input and select btw in the menu.

Source: Google Antigravity Changelog — describes PDF attachments through drag and drop or Add Context, and /btw as a side question to an ephemeral agent with context from the current conversation.

Then paste a short agent instruction like this:

Read the attached PDF and compare it with the project's README and open issues. Give me a short action list: which requirements are already covered, which are missing, and which two changes should we review first? Mark anything that depends on assumptions.

Good output should include:

• a clear split between document facts and assumptions
• references to relevant parts of the PDF or project context
• at most two prioritized next changes, not a long wish list
• a short note on the human review needed before anything changes

This is a good candidate for Hammer Automation's Tool Forge work: a small, controlled agent workflow where documents, code, and decision support connect without losing control. If you want to test the same pattern in your environment, start with a PDF that already affects a real decision and schedule the review before the agent changes anything: contact Hammer Automation.