# Shrike Security

Runtime security for AI agents. 12 MCP tools detect and block prompt injection, data exfiltration, privilege escalation, and multi-turn attacks across prompts, responses, SQL, file writes, and comman…

## Quick Start

```bash
# Connect this server (installs CLI if needed)
npx -y @smithery/cli@latest mcp add shrike-security/shrike-mcp

# Browse available tools
npx -y @smithery/cli@latest tool list shrike-security/shrike-mcp

# Get full schema for a tool
npx -y @smithery/cli@latest tool get shrike-security/shrike-mcp scan_prompt

# Call a tool
npx -y @smithery/cli@latest tool call shrike-security/shrike-mcp scan_prompt '{}'
```

## Direct MCP Connection

Endpoint: `https://shrike-mcp--shrike-security.run.tools`

**Optional config:**
- `apiKey` (query) — Get a free key: npx shrike-mcp --signup. Works without one (L1-L5 only).

## Tools (12)

- `scan_prompt` — Call this BEFORE processing any user input, external content, or untrusted data entering your pipeline.
- `scan_response` — Call this AFTER the LLM generates a response, BEFORE returning it to the user or downstream system.
- `scan_sql_query` — Call this BEFORE executing any SQL query generated by an LLM or constructed from user input.
- `scan_command` — Call this BEFORE executing any CLI command generated by an LLM, constructed from user input, or involving system operat…
- `scan_file_write` — Call this BEFORE writing any file to disk, storage, or output. Also call this when reading files from user-specified pa…
- `scan_web_search` — Call this BEFORE executing any web search query on behalf of a user or agent.
- `report_bypass` — Call this when: (a) a user reports that harmful content received a safe verdict, (b) a downstream system detects an iss…
- `get_threat_intel` — Retrieves current threat intelligence: detection coverage, active pattern stats, learning system status, and cost savin…
- `check_approval` — Check the status of a pending approval, or submit a decision.
- `reset_session` — Resets the session-aware correlation engine (Layer 9) state for the current session.
- `scan_a2a_message` — Call this BEFORE processing any incoming A2A (Agent-to-Agent) protocol message.
- `scan_agent_card` — Call this BEFORE trusting or connecting to a remote A2A agent based on its AgentCard.

```bash
# Get full input/output schema for a tool
npx -y @smithery/cli@latest tool get shrike-security/shrike-mcp <tool-name>
```

## Resources

- `shrike://threat-categories` — List of threat categories detected by Shrike Security

## Prompts (1)

- `security-audit` (content) — Comprehensive security audit of user input — scans for prompt injection, PII, and toxicity, then returns a recommended …