@benchspan/sdk

Prompt injection firewall for LLM agents. One-line integration that scans tool outputs and user messages before they reach your model.

Install

npm install @benchspan/sdk

Quick Start

import { BenchGuard } from "@benchspan/sdk";

const guard = new BenchGuard({ apiKey: "ag_live_...", agent: "my-agent" });

// Direct scan
const result = await guard.scan("some tool output", { role: "tool" });
// result.injection  -> true/false
// result.verdict    -> "block" | "warn" | "pass"

Framework Integrations

Vercel AI SDK

import { BenchGuard } from "@benchspan/sdk";
import { wrapLanguageModel, generateText } from "ai";
import { anthropic } from "@ai-sdk/anthropic";

const guard = new BenchGuard({ apiKey: "ag_live_..." });
const model = wrapLanguageModel({
    model: anthropic("claude-sonnet-4-6"),
    middleware: guard.asMiddleware(),
});
const { text } = await generateText({ model, prompt: "Hello" });

LangChain JS

import { BenchGuard } from "@benchspan/sdk";
import { ChatAnthropic } from "@langchain/anthropic";

const guard = new BenchGuard({ apiKey: "ag_live_..." });
const llm = new ChatAnthropic({ model: "claude-sonnet-4-6" });
const result = await llm.invoke(messages, { callbacks: [guard.asLangChainCallback()] });

OpenAI Agents SDK

import { BenchGuard } from "@benchspan/sdk";

const guard = new BenchGuard({ apiKey: "ag_live_..." });
const agent = new Agent({
    name: "assistant",
    model: "gpt-4o",
    hooks: guard.asAgentHooks(),
});

Google ADK

import { BenchGuard } from "@benchspan/sdk";
import { LlmAgent } from "@google/adk";

const guard = new BenchGuard({ apiKey: "ag_live_..." });
const agent = new LlmAgent({
    name: "assistant",
    model: "gemini-2.5-pro",
    beforeModelCallback: guard.asAdkCallback(),
});

Raw OpenAI / Anthropic SDK

import { BenchGuard } from "@benchspan/sdk";
import Anthropic from "@anthropic-ai/sdk";

const guard = new BenchGuard({ apiKey: "ag_live_..." });
const client = new Anthropic();

const result = await guard.wrapCall(messages, () =>
    client.messages.create({ model: "claude-sonnet-4-6", messages })
);

Modes

• "block" (default) -- Waits for scan result. Throws InjectionDetectedError if injection detected.
• "warn" -- Fires scan in the background with zero added latency. Logs a warning if injection detected but never blocks.

// Production -- block injections
const guard = new BenchGuard({ apiKey: "ag_live_...", mode: "block" });

// Evaluation -- monitor without blocking
const guard = new BenchGuard({ apiKey: "ag_live_...", mode: "warn" });

How It Works

1. Framework hook fires before each LLM call
2. SDK scans user and tool messages via the BenchSpan API
3. system and assistant messages are skipped (trusted)
4. Already-scanned messages are deduplicated across multi-turn conversations
5. In block mode, InjectionDetectedError is thrown if injection detected
6. In warn mode, the scan runs async -- zero latency added to your agent

Links

• Dashboard
• Docs