Integrate Arcjet Guard Tanstack AI

Integrate Arcjet Guard into TanStack AI

@arcjet/guard's TanStack AI v0 namespace wraps the agent's existing Arcjet client. It never talks to the Arcjet API itself. Two surfaces, one decision rule:

• Tool calls → guardMiddleware(). A chat({ middleware }) middleware whose onBeforeToolCall is the chat()-wide gate. Default DENY is { type: "skip", result: ArcjetDenialResult } so the tool never runs and the model sees the payload. Optional onDeny: "abort" returns { type: "abort", reason } and stops the run — the model does not get ArcjetDenialResult. Do not throw from the hook. Tools already branded by a sibling guardTool are skipped so Guard is not double-called. Inbound guard() before chat() does not brand tools and does not skip this gate.
• Correlation → tanstackAiContext() reads a caller-owned id from the helper options or chat({ context }). It never mints a new id. It never reads ctx.threadId (TanStack auto-generates it). It never reads traceId / requestId / streamId.

There is no guardTool. A throw from execute is swallowed into { error } and is not a usable deny envelope. There is no guardInbound, no guardApproval, and nothing named contentGuardMiddleware (TanStack already has that name).

This namespace is TanStack AI chat({ middleware }) + onBeforeToolCall. Not the Vercel AI SDK. Do not also wrap with @arcjet/guard/vercel-ai/v7 or @arcjet/guard/claude-managed-agents/v0. Client tools and provider-native tools with no local execute are out of scope.

Docs live at docs.arcjet.com/guards/tanstack-ai/. Do not overwrite any other /guards/... slug.

Put Arcjet first in the middleware array. onBeforeToolCall is first-win; if toolCacheMiddleware (or anything else) skips first, Guard never runs.

Screen inbound before chat() — there is no inbound hook.

There is no first-class inbound channel, so there is no guardInbound. Put prompt-injection (and other inbound rules) in the application before chat(). Call guard() directly. guard() fails open — callers must check hasFailedOpen(). TanStack's contentGuardMiddleware redacts the stream; it is not this policy gate.

needsApproval / defineInterrupt / onInterruptBoundary is HITL, not a policy gate.

needsApproval / defineInterrupt / onInterruptBoundary is human-in-the-loop. After a human yes, Guard still runs on the tool call. Same trap as Mastra requireApproval, Claude canUseTool, LangGraph interrupt(), Genkit toolApproval, OpenAI Agents needsApproval, LangChain humanInTheLoopMiddleware, and Google ADK requireConfirmation. There is no guardApproval.

Questions to ask the human first

Ask only what you cannot infer from the code; suggest defaults.

1. Which tools are risky (external side effects, irreversible, spends money, sends messages)? Those are gated by guardMiddleware.
2. What limits? (e.g. "10 lookups/min per order" → tokenBucket.)
3. Who is the user for metadata — an opaque user/tenant ID (never PII)? Default: none. Pass it via metadata on the policy. Put the conversation / session id you already have on chat({ context: { sessionId } }). That id is the correlation id. Do not use ctx.threadId.
4. Is an Arcjet outage unacceptable? Every helper defaults to onGuardError: "deny". Ask explicitly about inbound screening before chat(): failing closed there means the chat does not run for the duration of the outage, so "allow" is a routine and legitimate choice at that one call site. guard() itself still fails open — check hasFailedOpen().

The six things readers get wrong

1. There is no guardInbound. Screen prompt injection before chat() with guard(). Check hasFailedOpen(). contentGuardMiddleware is not Guard.
2. needsApproval / defineInterrupt / onInterruptBoundary is not a policy gate. It is HITL. After a human yes, Guard still runs.
3. The import path is versioned and there is no alias. @arcjet/guard/tanstack-ai/v0. @arcjet/guard/tanstack-ai does not resolve. There is no /v1 until TanStack AI ships 1.x. Docs are /guards/tanstack-ai/.
4. Correlation is read, never minted. Do not call createAgentContext inside a middleware callback — that generates a second id and splits the Sequence. Put the id you already chose on chat({ context }). Do not read ctx.threadId, traceId, requestId, or streamId.
5. Put Arcjet first. onBeforeToolCall is first-win. If toolCacheMiddleware skips first, Guard never runs.
6. Do not add guardTool and do not double-wrap with @arcjet/guard/vercel-ai/v7. TanStack AI is not the Vercel AI SDK. A throw from execute becomes { error }.

Step 1: Install and find the guard client

Install @arcjet/guard (required), plus @tanstack/ai (optional peer, needed for @arcjet/guard/tanstack-ai/v0). Always use the versioned path: @arcjet/guard/tanstack-ai/v0 resolves; @arcjet/guard/tanstack-ai throws ERR_PACKAGE_PATH_NOT_EXPORTED. The peer range is >=0.8.0 <1. Node 22+ — do not bump Node for this adapter.

npm install @arcjet/guard @tanstack/ai

If the agent has no guard client yet, launch one once at module scope:

import { launchArcjet } from "@arcjet/guard";

export const arcjet = launchArcjet({ key: process.env.ARCJET_KEY! });

Step 2: Gate tool calls — Arcjet first

import { chat } from "@tanstack/ai";
import { toolCacheMiddleware } from "@tanstack/ai/middlewares";
import { guardMiddleware } from "@arcjet/guard/tanstack-ai/v0";
import { tokenBucket, localDetectSensitiveInfo, policyInput } from "@arcjet/guard";

import { arcjet } from "./arcjet.js";

const lookupLimit = tokenBucket({
  refillRate: 10,
  intervalSeconds: 60,
  maxTokens: 10,
});
const detectPii = localDetectSensitiveInfo();

const stream = chat({
  adapter,
  messages,
  tools: [lookupOrder, ...mcpTools],
  context: { sessionId: conversationId },
  middleware: [
    guardMiddleware(arcjet, {
      action: ({ toolName }) => `${toolName}.invoked`,
      actor: conversationId,
      inputs: ({ toolName }) => ({
        tool: policyInput.server.string(toolName),
      }),
      rules: ({ toolName, input }) => {
        const note =
          typeof input === "object" && input !== null && "note" in input
            ? String((input as { note?: unknown }).note ?? "")
            : "";
        return [
          lookupLimit({ key: toolName, requested: 1 }),
          ...(note.length > 0 ? [detectPii(note)] : []),
        ];
      },
      sessionId: conversationId,
    }),
    toolCacheMiddleware(),
  ],
});

• Omit rules to submit none. The guard call still happens.
• Optional actor and inputs (static, or a resolver over this adapter's native call — parsed input plus trusted runtime/context) are forwarded on the guard call so a remote policy that declares those names can evaluate. Build each input with policyInput.
• On DENY the original execute never runs. Default delivery is { type: "skip", result: { arcjetDenied: true, reason, message, retryable } }.
• onDeny: "abort" stops the chat run with { type: "abort", reason } (the denial message string). The model does not get ArcjetDenialResult. Prefer default skip when the model should see the payload. onDeny: "abort" applies to real DENY only; unavailable stays skip.
• Default onGuardError: "deny" blocks the tool if Arcjet is unreachable.
• ALLOW captures outcome: "success" when the policy lets the tool run, not when execute finishes. onBeforeToolCall cannot wrap the tool; a later tool throw does not flip that capture.
• Tools already branded by a sibling guardTool skip the middleware so Guard is not double-called. This namespace has no guardTool. Inbound guard() before chat() does not stamp that brand.

Step 3: Screen inbound before chat

import { detectPromptInjection } from "@arcjet/guard";
import { tanstackAiContext } from "@arcjet/guard/tanstack-ai/v0";

import { arcjet } from "./arcjet.js";

const inbound = detectPromptInjection();
const decision = await arcjet.guard({
  label: "message.received",
  rules: [inbound(userText)],
  ...tanstackAiContext({ context: { sessionId: conversationId } }),
});

if (decision.conclusion === "DENY") {
  throw new Error("message blocked");
}
if (decision.hasFailedOpen()) {
  throw new Error("inbound screening failed open");
}

await chat({
  adapter,
  messages: [{ role: "user", content: userText }],
  context: { sessionId: conversationId },
  middleware: [guardMiddleware(arcjet, { sessionId: conversationId })],
});

There is no guardInbound. guard() fails open — always check hasFailedOpen().

Step 4: Correlation

Put the id you already have on chat({ context }):

await chat({
  adapter,
  messages,
  context: { sessionId: conversationId },
  middleware: [guardMiddleware(arcjet, { sessionId: conversationId })],
});

Preference order: context.correlationId, then context.sessionId, then context.conversationId, then init.sessionId / init.correlationId. If none is a valid 1–256 printable-ASCII string, the call is uncorrelated rather than joined to a generated id nobody has.

Pass a caller-owned bag as tanstackAiContext({ context: appContext }) or chat({ context: appContext }). A bare object that also has string requestId and streamId looks like TanStack's middleware envelope, so top-level sessionId on that object is ignored.

Never mint a new id. Never read ctx.threadId (TanStack auto-generates it). Never read traceId / requestId / streamId. needsApproval resumes after a human yes — Guard still runs on the tool call. Do not treat the interrupt or its resume value as correlation.

Verify the integration

1. npm run typecheck passes.
2. Exercise inbound PI (before chat(), including hasFailedOpen()), a middleware skip-deny, an abort-deny, abort+unavailable still skip, first-win ordering (Arcjet before toolCacheMiddleware), no-throw, never-mint, and fail-closed (an unreachable guard). Confirm the denial is { type: "skip", result } (or abort) and the run is not an interrupt.
3. Confirm in the Arcjet dashboard that decisions share the caller-owned session id as their correlation id — not threadId.
4. Manual E2E with a real ARCJET_KEY is still-to-verify until you run it.

A full working demo will land in arcjet/examples tanstack-agent as a later follow-up. Do not add an example under examples/ in the JS SDK repo.

Note: capture events are fire-and-forget and batched, so events can lag the decisions they accompany by a few seconds. A dropped event is diagnosed, never thrown.