AI SDK MCP Client — Watchlist

MCP100 Assessment

Watchlist / Tracking

Why we’re tracking it

Follow-up value is specific to MCP client negotiation and conversion into AI SDK tools: Does stateless discovery still fall back correctly for legacy stdio servers without leaking sessions?

Reviewed Oct 5, 2026 · Documentation review; complete setup workflow not independently tested

Best fit

TypeScript AI SDK applications that need to consume remote or local MCP tools.

What we know

@ai-sdk/mcp is an MCP client package for Vercel AI SDK. It connects to MCP servers and exposes their tools to AI SDK generation and agent APIs.

Maintenance / activity evidence

@ai-sdk/mcp 2.0.66 published 2026-10-01T19:20:32.073Z

Tracking evidence observed Oct 5, 2026. These dated signals do not establish current maintenance. Reviewed enrichment confidence: high.

Verification gaps / watch-outs

Installation and comprehensive security have not been independently audited.
Inherits all capabilities exposed by connected MCP servers. Writes: Depends on connected tools.
Depends on the target MCP server.
Sensitive access: Any data/credentials supplied to target servers

The library itself is an adapter, but it can connect an application to arbitrary MCP tool authority; review each target server separately.

Full authority, access and evidence limits are available in Review context and Access & risk.

Popularity

27,149 shared-repository stars · 17.6M package downloads / 30 days

Stars observed Oct 6, 2026. Downloads observed Oct 6, 2026.

Repository stars are shared with the full vercel/ai monorepo; package downloads are exact to @ai-sdk/mcp.

All limitations / Review context

Does stateless discovery still fall back correctly for legacy stdio servers without leaking sessions?

The bounded review of the publisher's documented MCP client negotiation and conversion into AI SDK tools and release notices identified no disclosed material condition preventing an accurately scoped public tracking entry. The resource's operational permissions remain uncertainty; this assessment does not certify installation safety.

The library itself is an adapter, but it can connect an application to arbitrary MCP tool authority; review each target server separately.
Inherits all capabilities exposed by connected MCP servers.
Depends on the target MCP server.
Requests and tool payloads are sent to configured MCP servers.
Reviewed write actions: Depends on connected tools
Sensitive access: Any data/credentials supplied to target servers
Installation, client compatibility and comprehensive security have not been independently audited. Authentication and access depend on the configured services.
Runtime behavior, dependencies and undisclosed vulnerabilities were not independently audited.
Documented tool authority and any remote data handling must be checked before installation or use.
Tool inventory, client matrix and permissions completeness were not requirements for this Radar tracking review.
This package is an MCP client, not a server.
The connected MCP server determines permissions, data handling and tool risk.

Original activity observations

@ai-sdk/mcp 2.0.66 published 2026-10-01T19:20:32.073Z

Technical identity

npm/@ai-sdk/mcp

Exact implementation source: https://www.npmjs.com/package/@ai-sdk/mcp ↗
Reviewed resource type: SDK
Repository identity: vercel/ai
Package identity: npm:@ai-sdk/mcp

Evidence & Sources

Evidence basis: Recorded source evidence
Setup runtime verification: Not independently runtime-tested by MCP100 for this setup workflow.

Reviewed Oct 5, 2026 · Latest source observation Oct 6, 2026

Package registry ↗
@ai-sdk/mcp
Recorded package identity, distribution or download evidence.
Includes references whose attribution is not established.
GitHub repository ↗
vercel/ai
Repository reference; implementation attribution is not established for every document.
Includes references whose attribution is not established.

Setup · Watchlist

SDK setup

Publisher instructions.

Source evidence observed Oct 14, 2026.

Additional setup details required — see the specific gaps below.

Install / connect
Publisher instructions
Publisher-documented install or connect path; prerequisites and any extra host configuration remain required.
sdk dependency install · Publisher instructions
npm install @ai-sdk/mcp
Configure / authenticate
MCP100 setup guide
This is a Vercel AI SDK MCP client integration package, not an independently launched MCP server. Install it in the application, configure its client/transport and auth in app code, then connect to a real remote or local MCP service.
Check connection
Suggested check
Use an SDK client connection and tool listing in the application; package installation alone does not create MCP tools.
Remove / uninstall / revoke
MCP100 setup guide
For the application dependency, run `npm uninstall @ai-sdk/mcp` in the owning Node project, remove integration imports and rebuild/redeploy. No standalone MCP registration exists for this package.
Client removal, software uninstall, persistent-data deletion and credential revocation are separate actions.
Package / application cleanup · matching installation environment · MCP100 composed guide
npm uninstall @ai-sdk/mcp

Official Documentation & Sources

https://www.npmjs.com/package/@ai-sdk/mcp ↗

What it does

@ai-sdk/mcp is an MCP client package for Vercel AI SDK. It connects to MCP servers and exposes their tools to AI SDK generation and agent APIs.

Capability: AI & Reasoning
Resource type: SDK
Industry: Software & Technology
Best fit: TypeScript AI SDK applications that need to consume remote or local MCP tools.

Representative uses

Connect Streamable HTTP MCP servers
Connect local stdio MCP servers
Pass bearer headers to remote MCP endpoints
Use MCP tools in generateText or streamText

Limitations

This package is an MCP client, not a server.
The connected MCP server determines permissions, data handling and tool risk.

Reviewed access

Access & risk

Review the exact access boundary before connecting this MCP to accounts, credentials, or sensitive data.

Authentication: Depends on the target MCP server.
Permissions: Inherits all capabilities exposed by connected MCP servers.
Data handling: Requests and tool payloads are sent to configured MCP servers.
Cost: unknown
Risk and cautions: The library itself is an adapter, but it can connect an application to arbitrary MCP tool authority; review each target server separately.

Access scope and cautions

Write actions: Depends on connected tools
Sensitive access: Any data/credentials supplied to target servers
Risk level: moderate

FAQ

Based on reviewed evidence as of 2026-10-06. Watchlist tracking is not a recommendation or a security certification.

Which tasks are documented for this implementation?

The reviewed tasks include: Connect Streamable HTTP MCP servers; Connect local stdio MCP servers; Pass bearer headers to remote MCP endpoints; Use MCP tools in generateText or streamText.

What setup and authentication are documented?

Recorded installation/entrypoint step: npm i @ai-sdk/mcp ai zod. Recorded transport: stdio, streamable_http. Prerequisites: Node.js/npm. These are the documented setup details, not a claim that installation or client compatibility was independently tested. Authentication: Remote-server specific; bearer headers can be configured.

Can it modify data or perform write operations?

Documented write actions include: Depends on connected tools. Recorded permission boundary: Inherits all capabilities exposed by connected MCP servers.

What data-handling boundary should I consider?

Requests and tool payloads are sent to configured MCP servers. Recorded sensitive-access context: Any data/credentials supplied to target servers.

What limitations and risks should I check before use?

The library itself is an adapter, but it can connect an application to arbitrary MCP tool authority; review each target server separately. This package is an MCP client, not a server. The connected MCP server determines permissions, data handling and tool risk. Runtime behavior, dependencies and undisclosed vulnerabilities were not independently audited. Documented tool authority and any remote data handling must be checked before installation or use.

Is this an official MCP?

Vercel is listed as First-party MCP in MCP100’s current evidence.