What Tablif read on Assay's site
Assay describes itself as: “Assay Execution-grounded security verifier for AI agent skills”
- Price model: free tier
- Free tier: Free forever
- Stage: Live, no waitlist
What it is
Calls itself “security verifier”.
- Calls itself: security verifier
“Execution-grounded security verifier for AI agent skills”
Where it puts itself
Rules out “unconfined mode”.
- Says it will not: unconfined mode
“There is no unconfined mode.”
What it costs
A free tier, free forever.
- Free plan: free forever
“Open source, v0.1.3.”
“Open source”
Who it sells to
Sells to developers.
- Sells to: developers
“Execution-grounded security verifier for AI agent skills”
How it runs
A command-line tool for Linux and macOS. Runs locally and works offline. Distributed through the pip. Open source. AI is the core of the product. You can bring your own API key.
- Kind: cli
- Runs on: Linux · macOS
- Runs where: on the device, offline
- Distributed via: pip
- Open source: yes
- AI role: core to the product
- Bring your own key: yes
“Install the CLI with uv or pip”
“Sandboxed checks require macOS or Linux.”
What it works with
Integrates with MCP. Analyzes / reviews MCP. Names pip and uv on its pages.
- Integrates with: MCP
- Analyzes / reviews: MCP
- Names: pip · uv
“Assay runs an AI agent skill in a sandbox on macOS or Linux against a real agent”
“MCP (separate capability) ( src/assay/sandbox/session.py , src/assay/mcp/ ) — a long-lived SandboxSession , a stdlib stdio MCP client”
What it offers as proof
Live, no waitlist. Claims “20-case”. Publishes docs.
- Number claimed: 20-case
“A 20-case benchmark”
“…docs -> https://github.com/haqaliz/assay/tree/master/docs”
Sources
Tablif read these pages, last on 17 Nov 2026. Everything above is what the product says about itself; Tablif does not verify every claim.
Similar products in Tablif Town
- AegisKernel: Zero-Trust AI Agent Security Runtime
- aiSecurityWarden: Human approval gate for your AI agent's secrets and actions
- Agentskills: Build faster with AI Agent Skills
- Agent Shield: Runtime security mesh for autonomous AI coding agents
- BLAZLE: In-flight security for AI agents
- WBX Sandbox: Secure execution for AI agents - Give your AI agents a safe place to run code
- NeuralTrust: Secure and govern enterprise AI agents with real-time identity-aware guard
- Claustra: Security for AI coding agents