An Antigravity alternative that stays open source and runs any model

Google Antigravity vs octomind

How to switch from Google Antigravity

1. 1

   Install octomind and keep Gemini on your own key

   One script installs the binary to ~/.local/bin on macOS or Linux. Create a key in Google AI Studio and pass any Gemini model id with google-studio:, or swap in another provider. -m sets the main model only; supervisor and compression calls stay on octohub:auto until you set [supervisor.model] and [compression.model], or run octomind login.

   curl -fsSL https://octomind.run/install.sh | bash
   
   export GOOGLE_STUDIO_API_KEY=...      # from Google AI Studio
   octomind run developer:general -m google-studio:<gemini-model-id>
   
   # or any other provider, or a model on your own machine
   # octomind run developer:general -m deepseek:deepseek-flash
   # octomind run developer:general -m local:qwen3-coder

2. 2

   Bring your rules, skills and MCP servers over

   Rename GEMINI.md to AGENTS.md. Antigravity reads AGENTS.md too, so both tools share one file. Skills in the repo under .agents/skills/ work as they are; copy global ones into ~/.config/agents/skills/. Move each server from mcp_config.json into a [[mcp.servers]] block in a TOML file in the octomind config directory.

   git mv GEMINI.md AGENTS.md
   mkdir -p ~/.config/agents/skills
   # global skills: the app and IDE keep them in ~/.gemini/config/skills,
   # the Antigravity CLI in ~/.gemini/antigravity-cli/skills
   cp -R ~/.gemini/config/skills/* ~/.config/agents/skills/
   
   # was in mcp_config.json: "playwright": { "command": "npx", "args": ["@playwright/mcp@latest"] }
   cat > ~/.local/share/octomind/config/mcp-playwright.toml <<'EOF'
   [[mcp.servers]]
   name = "playwright"
   type = "stdio"
   command = "npx"
   args = ["@playwright/mcp@latest"]
   timeout_seconds = 30
   tools = []
   auto_bind = ["developer:general"]
   EOF

3. 3

   Run your first job, with a spend stop

   Run octomind run developer:general in the repo for an interactive session. For scripts and CI, pipe the task in. The -c flag caps each request in US dollars, and a run that hits the cap exits with 75 so your script can tell a budget stop from a failure.

   cd ~/your/project
   octomind run developer:general -m google-studio:<gemini-model-id>      # interactive
   
   echo "The test suite fails. Find the bug, fix it, run the tests." \
     | octomind -c max_request_spending_threshold=0.50 \
         run developer:general -m google-studio:<gemini-model-id> --format plain
   echo "exit $?"                                                         # 75 = spend stop

Other options worth a look

• Gemini CLI with an API key↗

  Still Apache-2.0, and it still signs in with a Gemini API key or Vertex AI. Its README says it is not optimized for the latest Gemini models. It suits you if you liked the old CLI and can live without Google login.

• opencode↗

  An MIT-licensed agent for the terminal, desktop and IDE that connects 75+ providers through Models.dev, local models included. It suits you if you want open source and a wide provider list.

Where Google Antigravity is better

• No desktop app and no IDE extensions

  Antigravity has an app for running agents in parallel, its own IDE, and extensions for five editors. Octomind is a terminal agent. It speaks the Agent Client Protocol through octomind acp, and the docs show a Zed setup. Read more →

• No approval prompts, and the sandbox blocks writes only

  Tools run without asking. Guards in .agents/guardrails.toml block matching calls, and --sandbox limits writes on Linux and macOS. It does not cut the network, it is off by default, and it does nothing on Windows. Read more →

• No free model quota

  Antigravity's individual tier is $0 with weekly limits, Gemini included. With octomind, every call bills to your provider key, unless you run a local model and pay only for your hardware. Read more →

• GEMINI.md and folder-level rules are not read

  Octomind reads one AGENTS.md from the directory you start in. Antigravity merges rules from every folder up to the workspace root and from ~/.gemini/. Fold what you need into that one file.

Questions

More comparisons

• Claude Code alternative The same terminal agent loop on any model, open ones included, under Apache-2.0.
• OpenClaw alternative A personal agent you self-host as one binary, or hand to octos on Cloud
• opencode alternative Same open model as opencode: 24 of 25 tasks solved to its 19, at about half the cost.
• Hermes Agent alternative Lessons only from your corrections, as plain files you can edit, on your box or on Cloud
• Claude Cowork alternative Hand off the job once or daily, on any model; the result lands in Telegram, Slack or email
• Manus alternative Sourced reports and live pages from a cloud machine, on one balance that stops
• Codex alternative Keep your ChatGPT plan, add any other model, and stop a request at a dollar limit.
• Cursor alternative The agent without the editor: any model, your own keys, open source, next to any IDE.
• GitHub Copilot alternative The open-source agent for terminal and CI, on any model, billed by your own provider.
• DeepSeek Harness alternative DeepSeek models in a terminal and CI agent with tagged releases and a spend stop.

Reading

• Octomind vs Google Antigravity, side by side
• Gemini CLI alternatives after consumer access ended
• Self-improving agents: Harness² vs Octomind