Base44
Build and edit Base44 apps from an AI assistant, and query data your app's users submit.
Drive Vercel's v0 app builder: create chats, send prompts, approve the agent's plans and permissions, fetch previews.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The v0 MCP server lets Claude, ChatGPT or any MCP-compatible agent drive Vercel's v0 app builder. Anthropic's directory lists 8 tools that create and continue v0 chats, read their messages, resolve the v0 agent's pending questions and permission requests, and fetch a preview URL. Sign-in is OAuth with one mcp scope, and each generation spends credits.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
scopes_supported of exactly ["mcp"] on 2026-08-22. There is no read-only grant, so consent cannot separate listing your chats from creating new ones that spend money. The boundary a reader would want — between the routine read and the credit-spending write — does not exist at the grant.import-claude-design-from-url appears in Anthropic's listing and nowhere in v0's 1.3 MB documentation corpus. We could not establish what it does.resolveTask can approve a shell command. Its confirmed-permissions variant grants the v0 agent permission to run a tool, script or MCP call it proposed. The approval must match the pending request, but no human sits between your agent and that approval.rm -rf guard, which v0 describes as not a complete safety boundary.privacy: 'public', which v0 defines as indexable by search engines. Whether the MCP tool exposes it is unverified.Yes. Every generation spends credits from your v0 balance, and v0's Create Chat response returns a creditsCost object alongside token counts. v0's pricing page gives Free accounts $5 of monthly credits and a 7-message daily cap, Premium $20 and Plus $30 per user. When credits run out, v0 states that generation pauses.
Not through the tools listed. v0 documents a Deploy Chat endpoint that triggers a Vercel deployment, but the MCP guide names only seven backing endpoints and that is not one of them. Publishing to production is a Publish button in the v0 web interface. Treat this as the current listing's shape, not a guarantee.
No. v0's Get Preview URL endpoint returns a URL, a short-lived token and an expiry timestamp, and v0 states the preview is reached by sending that token in an x-v0-preview-token header. Browsers cannot attach a custom header to an iframe, so v0 documents proxying previews through your own backend instead.
More than a plan. v0's task schema has five variants, and one of them, confirmed-permissions, is documented as resolving a request where the v0 agent wants to execute a tool, shell command, script or MCP call. Calling resolveTask on that variant grants the v0 agent permission to run the command it asked about.
One scope, named mcp. v0's authorization server metadata advertised scopes_supported of exactly that single value on 2026-08-22. There is no read-only variant offered, so nothing at the OAuth grant separates listing your existing chats from creating new ones that spend credits.
The client is holding a stale tool list. v0's troubleshooting section says to restart or reconnect the MCP client so it refreshes the server's tool list, then check the v0 API v2 reference for the resource you expected. Reconnecting also fixes authentication failures after a token expires.
Remove the connection and add it again. v0 states that if you need to switch users or teams you should remove the existing v0 MCP connection in your client and connect to https://v0.app/api/mcp again. The account or team is chosen during the OAuth approval step, not per request.
On most plans, yes. v0's pricing comparison lists Data Opt-out as available only on Business and Enterprise, and its security page describes opting out of content being used for model training as an Enterprise feature. Free, Premium and Plus accounts have no documented opt-out.
No, they are two separate connectors with separate endpoints. Anthropic lists Vercel at mcp.vercel.com with seven read-only tools for inspecting projects, deployments and teams. v0 runs at v0.app/api/mcp and drives the app builder, so its tools generate code and spend credits.
.md; this is the documentation URL Anthropic's directory publishes. · retrieved 2026-08-22llms.txt (1.3 MB) and semantic sitemap — · (retrieved 2026-08-22). Both grepped for the undocumented tool name; zero hits. · retrieved 2026-08-22initialize to https://v0.app/api/mcp returned HTTP 401 with WWW-Authenticate: Bearer resource_metadata="https://v0.app/.well-known/oauth-protected-resource"; that descriptor returned JSON naming the resource and authorization server, and https://v0.app/.well-known/oauth-authorization-server returned scopes_supported of ["mcp"] (2026-08-22) · retrieved 2026-08-22robots.txt (retrieved 2026-08-22). No Disallow rules and no Content-Signal line; the sitemap is declared. · retrieved 2026-08-22Connect v0 once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.