What Tablif read on Open Agent Safety Platform's site
Open Agent Safety Platform describes itself as: “Open Agent Safety Platform”
- Stage: Live, no waitlist
What it is
Calls itself “platform”.
- Calls itself: platform
“…platform”
Who it sells to
Made for developers.
- Made for: developers
“…software letting developers set safeguards”
How it runs
Open source. AI is the core of the product.
- Open source: yes
- AI role: core to the product
“Open source platform to contain and monitor AI agents”
What it works with
Runs inside OpenShell. Names Hugging Face, OpenAI and 1 more on its pages.
- Runs inside: OpenShell
- Names: Hugging Face · OpenAI · Sentry
“Nvidia says the platform could have prevented OpenAI's Hugging Face incident”
“Nvidia released the Open Agent Safety Platform, software letting developers set safeguards to stop AI agents from breaking out of containment.”
What it offers as proof
Live, no waitlist. Shows customer logos. Claims “17,000 agents”.
- Customer logos: yes
- Number claimed: 17,000 agents
“Partners include Cisco, Microsoft, Oracle, CoreWeave, Dell, HPE, Leno”
“…over 17,000 agents attacked infrastructure for days and weeks.”
Sources
Tablif read these pages, last on 29 Sept 2026. Everything above is what the product says about itself; Tablif does not verify every claim.
Similar products in Tablif Town
- The open source runtime for AI agents : The open source runtime for building agentic IA on your data
- Caelis: Open-source runtime for long-running AI agent work
- Agentlas Science: A free platform for human with AI research collaboration
- Free AI Agent - No API Key on Device: Secure and transparent agents, adjustable and money saving.
- OpenCharm: Open-source body for the AI agent you already run
- AgentShelf AI Operations Platform: Put AI to work for your business without code
- Rebuno: Open-source runtime for AI agents in production
- PROMETHEUS: Safe AI Dev & Scan Security: Run AI agents locally. Every plugin scanned before execution