Injectshield
InjectShield scans untrusted text for prompt injection patterns before it reaches your LLM's context window. The MCP server exposes three tools: scan for raw text, scan_url for remote content, and patterns to inspect the current ruleset. It runs heuristic detection across nine categories including instruction injection, role hijack, exfiltration attempts, and encoding smuggling, then returns risk scores and sanitized output. You'd reach for this when your agent ingests git commits, documentation, user inputs, or any external content you don't control. The detection patterns are open source MIT, while the hosted API at api.injectshield.dev adds metering, custom patterns, and semantic classification via Workers AI.
- tools
- 3
- auth
- Secret needed
- transport
- STDIO
- env vars
- 2
- last updated
- May 6, 2026
Tools
3 toolsPublic tool metadata for what this MCP can expose to an agent.
scanScan text for prompt-injection patterns before it reaches the agent's context.
Scan text for prompt-injection patterns before it reaches the agent's context.
No parameter schema in public metadata yet.
scan_url
No parameter schema in public metadata yet.
patterns
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageConfiguration
INJECTSHIELD_API_KEY*secretAPI key from https://injectshield.dev (free tier: 10K req/mo).
INJECTSHIELD_API_BASEdefault: https://api.injectshield.devOverride the API base URL for self-hosted deployments.
Readme
InjectShield
Prompt-injection firewall for AI agents.
A drop-in REST API that detects and neutralizes injection attacks in any text — git commits, web pages, files, emails, user inputs — before they reach your AI agent's context window.
This repo is the open-source heuristic ruleset plus the source for the managed API at promptshield.pages.dev.
Why
In May 2026 a viral HN thread demonstrated that a single git commit message could burn a Claude Code user's entire session quota via a schema-driven attack ("OpenClaw"). The pattern is general: any AI agent that ingests untrusted text — code review bots, documentation summarizers, RAG agents, support copilots — is exposed to prompt injection. Most teams ship without any input-side defense.
InjectShield is one layer of a defense-in-depth strategy. It's not a silver bullet. Use it alongside system-prompt hardening, tool sandboxing, and output filtering.
Install as an MCP (Claude Code, Cursor, Cline, ...)
Details
- Repository
- github.com/bch1212/injectshield
- Website
- injectshield.dev
- Categories
- AI & LLM Tools, Security & Pentesting
- Package
@injectshield/mcp- Transport
- STDIO
- Auth
- Required
- Registry
- MCP registry, active
More from bch1212
Related AI & LLM Tools MCP servers
View all
Omni Mcp Serverbitofacoder/omni-mcp-serverAll-in-one MCP server: GitHub, Git, Slack, web fetch, memory, and filesystem. Agent Mode opt-in.0stars
Vectorabrunosrz/vectoraAI assistant with RAG, web search, filesystem and memory. MCP sub-agent for Claude Code.0stars
CaffeinecaffeineCaffeine's MCP server lets you build full-stack apps from inside Claude, ChatGPT, Cursor, or other agents. Send a build prompt, get progress updates, and have y0stars
LLM SEO MCP — Elephant Accountabilitychris-eaccountability/elephant-accountability-mcpLLM SEO and Agent Discoverability for B2B SaaS. Pricing, fit assessment, audit requests.0stars
Agentmemorymcp Mcpclauxel/agentmemory-mesh-mcpA paid remote MCP for agent memory MCP, built to return verdicts, receipts, usage logs, and audit-re0stars
Mcp Serveryuz0101/gammainfra-mcp-serverSmart LLM routing across every major provider via one OpenAI-shape API.0stars