@rnsk/tools

Runstack meta tools for the Vercel AI SDK — install and drop 600+ toolkit tools into generateText, streamText, or agent loops.

Source: https://runstack.engineer/docs/rnsk-tools

@rnsk/tools wraps the Runstack MCP server as native Vercel AI SDK tools. One import gives your model the same MCP tools as the hosted MCP server: discover tools, check auth, connect OAuth, and execute — without stuffing hundreds of tool schemas into context.

Install

npm install @rnsk/tools ai zod

Create an API key (rnsk_…) in the Runstack app and store it in an environment variable (for example RUNSTACK_API_KEY). See API Keys.

Quick start

await generateText({
  model: "anthropic/claude-sonnet-4-5",
  prompt: "Star the vercel/ai repo on GitHub and email me a confirmation.",
  tools: runstackTools(process.env.RUNSTACK_API_KEY!),
})

Use the same tools object with streamText, ToolLoopAgent, or any AI SDK entry point that accepts a tool map.

Scope to one or more toolkits

When the agent only needs specific products, pass toolkit ids (same field as @rnsk/toolkits manifests). All four meta tools are still returned; search_tools results are filtered to those toolkits.

// One toolkit
tools: runstackTools("github", process.env.RUNSTACK_API_KEY!)

// Multiple toolkits
tools: runstackTools(process.env.RUNSTACK_API_KEY!, {
  toolkitIds: ["github", "gmail"],
})

Find the id on each toolkit docs page — it appears as a badge next to category, auth type, and tool count (for example web-search on the Firecrawl page).

Scope to Studio agents

Limit tool discovery to the toolkits in a Studio agent workflow by passing its agent id (copy from the agent card in Studio):

// One agent
tools: runstackTools(process.env.RUNSTACK_API_KEY!, {
  agentIds: "paste-agent-id-here",
})

// Multiple agents
tools: runstackTools(process.env.RUNSTACK_API_KEY!, {
  agentIds: ["agent-id-1", "agent-id-2"],
})

Use the agent's role as your system prompt:

const agent = await fetchRunstackAgent("paste-agent-id-here", process.env.RUNSTACK_API_KEY!)

await generateText({
  model: "anthropic/claude-sonnet-4-5",
  system: buildAgentSystemPrompt(agent),
  prompt: "Your task here",
  tools: runstackTools(process.env.RUNSTACK_API_KEY!, { agentIds: agent.id }),
})

See Runstack Studio for building agents and finding ids.

Merge with your own tools

const tools = {
  ...runstackTools(process.env.RUNSTACK_API_KEY!),
  myCustomTool: tool({
    /* ... */
  }),
}

What you get

Details: MCP tools.

Related