Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
69 changes: 69 additions & 0 deletions LAUNCHGUIDE.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,69 @@
---
name: agentshield
title: AgentShield
tagline: Zero-leak secret defense, prompt injection blocker & kernel sandbox for AI agents
description: Enterprise-grade runtime security, credential interception, and kernel sandboxing MCP proxy for Claude Desktop, Cursor, and autonomous AI agents. Blocks 60+ secret formats (API keys, SSH keys, database credentials), neutralizes prompt injection payloads, and enforces OS isolation (Linux Landlock, macOS Seatbelt, Windows AppContainer).
category: Security
use_cases:
- Secret Leak Prevention
- Prompt Injection Defense
- OS Kernel Sandboxing
- AI Security Proxy & Redactor
- CI/CD Security Compliance
getting_started:
- "Scan text for secrets and credentials: agentshield_scan text: 'export API_KEY=test-sample-token'"
- "Verify local cryptographic audit ledger: agentshield_verify_ledger"
- "Inspect host OS sandbox readiness: agentshield_doctor"
- "Tool: agentshield_scan: Scans prompt text or tool outputs for 60+ credential patterns and high-entropy secrets"
- "Tool: agentshield_verify_ledger: Verifies HMAC-SHA256 and Ed25519 tamper-evident audit ledger"
- "Tool: agentshield_doctor: Inspects active OS kernel sandbox mechanisms (Landlock, Seatbelt, AppContainer)"
tags:
- security
- ai-agents
- dlp
- prompt-injection
- credential-defense
- mcp
- sandboxing
- cursor
- claude
documentation_url: https://ritualdev.in/agentshield
repository: https://github.com/RitualDev-Lab/AgentShield
homepage: https://ritualdev.in/agentshield
npm: "@ritualdev-lab/agentshield"
pypi: agentshield-defense
transport: stdio
command: npx
args: ["-y", "@ritualdev-lab/agentshield", "mcp"]
pricing:
type: freemium
license: Apache-2.0
environment_variables: []
---

# AgentShield MCP Security Server

**AgentShield** provides client-side runtime security, credential interception, and kernel isolation for AI coding assistants and autonomous agents.

## Setup Requirements

**None.** AgentShield runs 100% locally with zero required external API keys or cloud tokens.

## Quick Installation

```json
{
"mcpServers": {
"agentshield": {
"command": "npx",
"args": ["-y", "@ritualdev-lab/agentshield", "mcp"]
}
}
}
```

## Exposed Tools

1. `agentshield_scan`: Scan prompt text or tool outputs for 60+ sensitive credential formats (AWS keys, OpenAI tokens, SSH keys) and high-entropy secrets.
2. `agentshield_verify_ledger`: Verify the local HMAC-SHA256 and Ed25519 tamper-evident audit ledger for compliance.
3. `agentshield_doctor`: Inspect host OS kernel sandbox capabilities (Linux Landlock/bwrap, macOS Seatbelt, Windows AppContainer) and security readiness.
51 changes: 14 additions & 37 deletions benchmarks/run-10k-matrix.ts
Original file line number Diff line number Diff line change
Expand Up @@ -144,51 +144,27 @@ export async function runBenchmarkMatrix(
const results: BenchmarkMetric[] = [];
try {
results.push(
await measureScenario(
"core_redactor_mixed",
iterations,
warmup,
maxP99Ms,
(i) => {
redactText(i % 2 === 0 ? cleanPayload : secretPayload);
},
),
await measureScenario("core_redactor_mixed", iterations, warmup, maxP99Ms, (i) => {
redactText(i % 2 === 0 ? cleanPayload : secretPayload);
}),
);

results.push(
await measureScenario(
"engine_scan_and_redact",
iterations,
warmup,
maxP99Ms,
(i) => {
engine.scanAndRedact(i % 2 === 0 ? cleanPayload : secretPayload);
},
),
await measureScenario("engine_scan_and_redact", iterations, warmup, maxP99Ms, (i) => {
engine.scanAndRedact(i % 2 === 0 ? cleanPayload : secretPayload);
}),
);

results.push(
await measureScenario(
"sandbox_command_inspect",
iterations,
warmup,
maxP99Ms,
(i) => {
sandbox.inspectCommand(i % 2 === 0 ? "git status --short" : "cat .env && printenv");
},
),
await measureScenario("sandbox_command_inspect", iterations, warmup, maxP99Ms, (i) => {
sandbox.inspectCommand(i % 2 === 0 ? "git status --short" : "cat .env && printenv");
}),
);

results.push(
await measureScenario(
"zygote_pool_dispatch",
iterations,
warmup,
maxP99Ms,
async () => {
await zygotePool.dispatch({ action: "inspect", command: "ls -la" });
},
),
await measureScenario("zygote_pool_dispatch", iterations, warmup, maxP99Ms, async () => {
await zygotePool.dispatch({ action: "inspect", command: "ls -la" });
}),
);
} finally {
zygotePool.shutdown();
Expand Down Expand Up @@ -226,7 +202,8 @@ export async function runBenchmarkMatrix(
}

const isMain =
process.argv[1] !== undefined && resolve(process.argv[1]) === resolve(fileURLToPath(import.meta.url));
process.argv[1] !== undefined &&
resolve(process.argv[1]) === resolve(fileURLToPath(import.meta.url));

if (isMain) {
const args = process.argv.slice(2);
Expand Down
4 changes: 4 additions & 0 deletions package.json
Original file line number Diff line number Diff line change
Expand Up @@ -64,5 +64,9 @@
"tsup": "^8.3.5",
"tsx": "^4.19.2",
"typescript": "^7.0.2"
},
"dependencies": {
"@modelcontextprotocol/sdk": "^1.32.2",
"zod": "^4.6.7"
}
}
Loading
Loading