Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.corsair.dev/llms.txt

Use this file to discover all available pages before exploring further.

You can use the Vercel AI SDK with Corsair through the dashboard or in code.

​
In the dashboard

  1. Sign in at app.corsair.dev.
  2. Open Dashboard and click New instance.
  3. Open your instance and click Add Plugin to enable an integration (e.g. GitHub, Slack).
  4. Click New User to add a tenant — enter a tenant ID and click Save.
  5. Connect the integration for that tenant: click the cell where the tenant row meets the plugin column, then sign in with OAuth or enter credentials.
  6. Click the tenant name in the first column to open the side panel.
  7. Copy the MCP URL and create an API key (New key → name → Create → copy the secret once).
For SDK apps you can also use a developer API key from API Keys in the sidebar (scoped to your instance) with tenant.mcp.createVercelClient().
Quick Start under SDK → Vercel copies a full starter if you want a guided demo.

​
In code

​
With developer API key (backend)

Uses your developer key + tenant.mcp.createVercelClient() — no separate MCP secret file: 
npm install @corsair-dev/app ai @ai-sdk/mcp @ai-sdk/openai

import { openai } from "@ai-sdk/openai";
import { createClient } from "@corsair-dev/app";
import { stepCountIs, streamText } from "ai";

const corsair = createClient({ apiKey: process.env.CORSAIR_DEV_KEY! });
const mcpClient = await corsair
  .instance(process.env.CORSAIR_INSTANCE_ID!)
  .tenant("alice")
  .mcp.createVercelClient();

const stream = streamText({
  model: openai("gpt-4.1"),
  tools: await mcpClient.tools(),
  stopWhen: stepCountIs(15),
  prompt: "Star the corsairdev/corsair repo on GitHub",
});

for await (const text of stream.textStream) {
  process.stdout.write(text);
}

await mcpClient.close?.();

​
With tenant MCP key

import { createClient, createVercelAiMcpClient } from "@corsair-dev/app";

const key = await corsair
  .instance(instanceId)
  .tenant("alice")
  .mcpKeys.create("vercel-agent");

const mcpClient = await createVercelAiMcpClient({
  url: key.mcpHttpUrl,
  apiKey: key.secret,
});

const tools = await mcpClient.tools();
Always await mcpClient.tools() before passing to streamText.

​
Run it

Run your script or pass any prompt to streamText. If you connected GitHub, the Quick Start prompt is: 
Use Corsair to star the Corsair repo on GitHub