AgentSeal

Spec sheet

What it does

AgentSeal is an AI-agent security toolkit that red-teams prompts, scans agent files and MCP configurations, and monitors changes.Read 15 Oct 2026 · agentseal.org

Purpose

AgentSeal is an open-source security scanner and toolkit for testing AI agent prompts, auditing MCP servers, and detecting agent vulnerabilities.Read 30 Sep 2026 · agentseal.org

Privacy

The privacy policy says prompts are processed in real time and not retained, while scan results and metadata are kept for account history.Read 15 Oct 2026 · agentseal.org

Support

The pricing page lists community support for Free and priority support for Pro; the contact page says replies typically arrive within 24 hours.Read 15 Oct 2026 · agentseal.org

License

The GitHub repository and PyPI metadata identify the license as FSL-1.1-Apache-2.0.Read 15 Oct 2026 · github.com

API and SDKs

AgentSeal documents a Python AgentValidator API, a TypeScript AgentValidator API, and a REST dashboard API.Read 15 Oct 2026 · github.com

Attack coverage

Pro adds 45 MCP-poisoning, 28 RAG-poisoning, and 13 multimodal probes plus behavioral genome mapping.Read 15 Oct 2026 · agentseal.org

CI integrations

AgentSeal supports SARIF 2.1.0, JUnit XML, GitHub Actions summaries, and policy-as-code rules for gating pull requests.Read 30 Sep 2026 · agentseal.org

CI/CD outputs

The CLI can emit terminal, JSON, SARIF, and JUnit results and enforce score or severity gates.Read 15 Oct 2026 · agentseal.org

Dashboard

The web dashboard provides scan history, trust-score trends, encrypted prompt storage, agent monitoring, API-key management, and GitHub integration.Read 15 Oct 2026 · agentseal.org

Deterministic results

The scanner uses hardcoded probes, canaries, and n-gram matching rather than an LLM judge, so repeated runs are reproducible.Read 15 Oct 2026 · agentseal.org

Independent license concern

Hysen Labs notes that FSL-1.1 is not OSI-approved and says the repository badge and metadata disagree, so users should read the license file before commercial use.Read 15 Oct 2026 · hysenlabs.com

Independent project summary

SourcePulse describes AgentSeal as a toolkit for dangerous-skill detection, poisoned-config scanning, prompt red-teaming, and live-system auditing.Read 15 Oct 2026 · sourcepulse.org

Installation

The project publishes installation through pip and npm packages.Read 15 Oct 2026 · github.com

License discrepancy

The official documentation describes the core scanner as MIT licensed, while the repository identifies FSL-1.1-Apache-2.0.Read 15 Oct 2026 · agentseal.org

Limit: adoption

RightAIChoice says community adoption is limited, making real-world effectiveness difficult to judge.Read 15 Oct 2026 · rightaichoice.com

Limit: dashboard bug

The public GitHub issue tracker shows an open bug titled inability to connect to the dashboard.Read 15 Oct 2026 · github.com

Limit: deployment

RightAIChoice says there is no plug-and-play cloud version and users may need to run the CLI or integrate it themselves.Read 15 Oct 2026 · rightaichoice.com

Limit: model costs

RightAIChoice notes that prompt scanning can require paid API access to an external language-model provider.Read 15 Oct 2026 · rightaichoice.com

LLM providers

The site lists Ollama, OpenAI, Anthropic Claude, OpenRouter, Google Gemini, DeepSeek, Groq, Together AI, LM Studio, llama.cpp, vLLM, and compatible OpenAI chat API endpoints.Read 30 Sep 2026 · agentseal.org

Machine guard

Guard scans local agent skill files, MCP configurations, hidden payloads, and toxic data flows without an API key.Read 15 Oct 2026 · agentseal.org

Machine protection

Guard scans agent configurations and skill files on a machine, while Watch monitors those files and MCP configs for changes.Read 30 Sep 2026 · agentseal.org

Maker identity

LinkedIn identifies the organization name as AgentSeal and categorizes it under Technology, Information and Internet.Read 15 Oct 2026 · linkedin.com

Maker size

LinkedIn lists AgentSeal as self-employed, founded in 2026, with 2–10 employees.Read 15 Oct 2026 · linkedin.com

MCP scanning

Scan-mcp connects to running MCP servers and analyzes tool definitions for poisoning, toxic flows, and rug pulls.Read 15 Oct 2026 · agentseal.org

Offline use

The CLI can run fully offline with a local Ollama model, with zero network calls and zero telemetry, according to the site.Read 30 Sep 2026 · agentseal.org

Praise: coverage

RightAIChoice reports praise for the open-source model, broad prompt/MCP/skill-file coverage, and multiple attack surfaces.Read 15 Oct 2026 · rightaichoice.com

Prompt scanning

The scan command tests system prompts with 225 base probes covering extraction and injection attacks.Read 15 Oct 2026 · agentseal.org

Prompt testing

Its scan command runs adversarial probes against system prompts to detect extraction and injection vulnerabilities.Read 30 Sep 2026 · agentseal.org

Providers

Documented providers include OpenAI, Anthropic, Ollama, LiteLLM, MiniMax, and arbitrary HTTP endpoints.Read 15 Oct 2026 · github.com

Python package

PyPI lists AgentSeal 0.10.0, requires Python 3.10 or newer, and classifies it as beta software.Read 15 Oct 2026 · pypi.org

Rating on iTechGuides

9.2 out of 10 from 0 reader reviews; the score is an editor rating.Read 15 Oct 2026 · itechguides.com

Real-time shield

Shield watches skill files and MCP configurations for changes and sends desktop alerts.Read 15 Oct 2026 · agentseal.org

Requirements

The installation page requires Python 3.10 or higher for the CLI, and says the npm wrapper shells out to the Python CLI.Read 30 Sep 2026 · agentseal.org

Research news

AgentSeal reported on March 28, 2026 that it runtime-tested six high-profile MCP servers and successfully exploited every flagged vulnerability.Read 15 Oct 2026 · agentseal.org

Scan coverage

The FAQ describes 311 probes across extraction, injection, MCP tool poisoning, RAG poisoning, and multimodal attacks.Read 30 Sep 2026 · agentseal.org

Security registry

The MCP registry page says each server passes through a seven-stage security pipeline before it is listed, including sandboxing and probing.Read 30 Sep 2026 · agentseal.org

Security storage

Dashboard prompts are stored with Fernet encryption described as AES-256, and local configuration files use owner-only permissions.Read 15 Oct 2026 · agentseal.org

System requirements

The repository states Python 3.10+ or Node.js 18+ is required.Read 15 Oct 2026 · github.com

Plans and prices

Every AgentSeal plan and price →

Where Specifiction places it

• AI Red Teaming Tools#1 of 279.4