Best MCP Servers for Google Antigravity in 2026
Firecrawl, Bright Data, Zyte, and Crawlora compared as Google Antigravity MCP servers: config, the serverUrl gotcha, and which fits which job.
Google Antigravity is growing fast enough that Firecrawl and Bright Data have each shipped their own setup guide for it — but neither has published the head-to-head comparison a buyer actually wants. This one runs that comparison, sourced from Antigravity's own docs and each vendor's own guide, not paraphrased from a listicle.
Antigravity's MCP config
Antigravity reads MCP server definitions from a JSON file at ~/.gemini/config/mcp_config.json (global) or .agents/mcp_config.json at your project root (workspace-scoped, takes precedence over global). The root key is mcpServers, and the same file format applies whether you're configuring Antigravity 2.0, the Antigravity IDE, or the Antigravity CLI.
Antigravity supports two server shapes:
- Command-based (local):
{ "command": "...", "args": [...], "env": {...}, "cwd": "..." }— runs a local process. - Remote:
{ "serverUrl": "...", "headers": {...} }— points at a hosted endpoint. Both shapes also acceptdisabled(boolean) anddisabledTools(array) toggles, and the remote shape additionally supportsauthProviderType: "google_credentials"for ADC or anoauthblock (clientId/clientSecret).
The gotcha worth flagging up front: Antigravity's config directory is named ~/.gemini/, the same prefix Gemini CLI uses, but the two schemas aren't identical. Gemini CLI splits remote servers into url (SSE) vs httpUrl (HTTP-streaming) — Antigravity's own docs state plainly that "legacy fields like url or httpUrl are not supported," using serverUrl for every remote server instead. Copy a Gemini CLI remote-server block into Antigravity as-is and it won't connect.
You don't have to hand-edit the file, either: an MCP Store handles installs through a UI — Antigravity 2.0: Settings → Customizations → Installed MCP Servers; Antigravity IDE: the Agent side panel's … menu → MCP Servers; Antigravity CLI: the /mcp command opens an interactive MCP manager.
What to evaluate, in Antigravity specifically
- Remote field name: Antigravity uses serverUrl for every remote server and rejects the url and httpUrl fields Gemini CLI uses, even though both live under ~/.gemini/. A block copied from Gemini CLI will not connect here.
- Keyless versus keyed: Firecrawl's hosted endpoint works with no key but shares a rate limit with everyone on your public IP and exposes three tools. Decide whether zero-setup or the full tool set matters more for the session.
- Local process or none: Bright Data runs through npx; Firecrawl's hosted endpoint and Crawlora are serverUrl-based with nothing to run. Antigravity's disabledTools array lets you trim either kind.
- Generic crawl (any URL → markdown) versus structured platform data (named sites → JSON). Plan for two servers, not one.
- Billing on failure: a blocked fetch inside an agent loop retries. Whether each attempt is billed changes the cost of a long session.
The contenders
| Tool | Format | Coverage | Config | Best for |
|---|---|---|---|---|
| Firecrawl MCP | Remote (hosted, keyless by default) | Scrape/search/parse any URL, 3 tools in keyless mode | serverUrl-based, https://mcp.firecrawl.dev/v2/mcp, no key required (OAuth or API key unlocks more) | Zero-setup crawl-to-markdown for arbitrary sites |
| Bright Data | Local MCP server (70+ tools) | 40+ named platforms + proxy/anti-bot infra | command-based, npx @brightdata/mcp, API_TOKEN (+ PRO_MODE: "true" for the full tool set) in env | Anti-bot-hardened extraction at scale |
| Zyte | Hosted remote MCP server | Zyte API and Scrapy Cloud tools | https://mcp.zyte.com/v1/mcp; Antigravity undocumented | Zyte API users |
| Crawlora | Remote MCP server | 3,635 tools across 459 platform groups | serverUrl-based, https://mcp.crawlora.net/mcp, Authorization: Bearer <key> under headers | Structured JSON from named platforms, pay-on-success, full tool set with no local process |
Firecrawl MCP: zero setup, three tools, a shared rate limit
Firecrawl's own Antigravity quickstart points at its hosted endpoint: serverUrl: "https://mcp.firecrawl.dev/v2/mcp" and nothing else. That gives you firecrawl_search, firecrawl_scrape and firecrawl_parse with no API key, which is the fastest MCP setup in this series. The trade-offs are that keyless access shares its rate limit with everyone on your public IP, and that it is a narrower slice than the full local npx -y firecrawl-mcp server used in other clients; adding an API key or OAuth unlocks more. Choose it when the job is occasional "read this page into context" and setup time matters more than tool breadth.
Bright Data: the only local process in this comparison
Bright Data publishes its own Antigravity guide with a command-based entry: npx @brightdata/mcp, API_TOKEN under env, and "PRO_MODE": "true" to unlock the full 70+-tool set (the default is the smaller Rapid tier). Because everything runs through a local npx process, Antigravity's disabledTools array is worth using to trim the browser-automation tools a coding agent rarely needs. Choose it when anti-bot reliability on hard, named targets is the constraint and running a process is acceptable.
Zyte: a hosted server now exists, Antigravity isn't documented
Update, October 5, 2026: Zyte launched a hosted MCP server on September 30, 2026 (https://mcp.zyte.com/v1/mcp), which this post first missed. Zyte's docs cover other clients, not Antigravity, so it is untested here. Antigravity expects remote servers under serverUrl, so Zyte's generic url JSON would need adapting. Tools and billing: see our Pi write-up.
Crawlora: serverUrl, full tool set, nothing to run
Crawlora's MCP server connects to Antigravity the same way Firecrawl's default does, a serverUrl-based remote entry, but at the full tool count and with your own key rather than a shared keyless pool. Point serverUrl at https://mcp.crawlora.net/mcp with Authorization: Bearer <key> under headers:
{
"mcpServers": {
"crawlora": {
"serverUrl": "https://mcp.crawlora.net/mcp",
"headers": { "Authorization": "Bearer <CRAWLORA_API_KEY>" }
}
}
}
Do not use url or httpUrl here even though the directory is ~/.gemini/; Antigravity's docs state those legacy fields are not supported. The 3,635 tools map one-to-one onto documented REST endpoints across search, maps, e-commerce, social, finance, travel, jobs, media, app stores and reviews; use disabledTools to scope a project to the handful it needs. Responses are normalized JSON, billed only on success, with 2,000 free credits a month and no card. Choose it when the agent needs a record from a named platform (a price, a ranking, a listing, a filing) rather than a page to read.
How to choose, by job
- "Read this page / these docs into context, right now, no setup" → Firecrawl MCP's keyless default; add a key if the shared rate limit bites.
- "Get me the data off this hard, named site, at volume" → Bright Data, with
disabledToolstrimming the browser tools. - "Give me the price / ranking / listing / filing from a known platform, as JSON, with nothing installed" → Crawlora MCP on
serverUrl. - "I want Zyte's Scrapy-generation workflow" → Claude Code or GitHub Copilot for now, not Antigravity.
- Config copied from Gemini CLI won't connect → replace
urlorhttpUrlwithserverUrl; it is the most common Antigravity MCP mistake.
Add structured web data to Antigravity
3,635 tools across 459 platform groups, one API key, billed only on success. 2,000 free credits a month, no card.
Sources
Related reading
- Best MCP Servers for Gemini CLI in 2026: the closely-related client with a different remote-server field — read this if you're deciding between the two.
- Best MCP Servers for Warp in 2026 · Best MCP Servers for JetBrains IDEs in 2026.
- Best MCP Servers for GitHub Copilot in 2026 · Best MCP Servers for Windsurf in 2026.
- Best Web Scraping MCP Servers & Skills for Claude Code in 2026.
- Crawlora vs Firecrawl · Crawlora vs Bright Data.
- Give your AI agent live web data with MCP.
Frequently asked questions
What's the best MCP server for web scraping with Google Antigravity?
It depends on the job. Firecrawl MCP's hosted keyless default fits occasional, zero-setup crawling of arbitrary sites (3 tools, shared rate limit). Bright Data is built for anti-bot-hardened extraction across 40+ named platforms and publishes its own Antigravity guide. Crawlora fits when you want structured JSON from named platforms at the full tool count without maintaining a parser or running a local process, billed only on success. Zyte launched an official hosted MCP server on September 30, 2026, but does not document Antigravity as a client.
Where does Google Antigravity read its MCP server config from?
A JSON file at ~/.gemini/config/mcp_config.json (global) or .agents/mcp_config.json (workspace-scoped, takes precedence), root key mcpServers — the same file format across Antigravity 2.0, the IDE, and the CLI. There's also a UI path with no JSON to edit: an MCP Store reachable from Settings → Customizations → Installed MCP Servers (2.0), the Agent panel's … menu → MCP Servers (IDE), or the /mcp command (CLI).
Can I copy a Gemini CLI MCP config into Google Antigravity?
Not directly for a remote server. Antigravity's config directory shares the ~/.gemini/ prefix with Gemini CLI, but the schemas differ: Gemini CLI splits remote servers into url (SSE) or httpUrl (HTTP-streaming), while Antigravity's own docs say those legacy fields aren't supported and use serverUrl instead for every remote server. A command-based (local) server config is the same shape in both, but a remote block needs its field renamed.
Does Zyte work with Google Antigravity?
Zyte launched an official hosted MCP server on September 30, 2026 (https://mcp.zyte.com/v1/mcp, account sign-in, billed at Zyte API rates with no separate MCP fee). Zyte's docs list Claude Code, Codex CLI, GitHub Copilot CLI, VS Code and claude.ai as supported clients, not Antigravity, so using it here is the generic remote-server path, which we have not tested. Access needs an organization with an active Zyte API or Scrapy Cloud plan.
How do I install Crawlora's MCP server for Google Antigravity?
Add a serverUrl-based entry pointing at https://mcp.crawlora.net/mcp, with the API key under a headers.Authorization field (Bearer <key>). No local process to install or run, and no keyless tool-count limit like Firecrawl's default. Free tier is 2,000 credits/month, no card.