To all AI builders and security pros — we’re excited to roll out a powerful new upgrade to Agentic Radar that brings even more transparency and visibility into your agentic workflows.
Agentic Radar now supports automatic detection of Model Context Protocol (MCP) server connections, giving teams deeper insight into how agents access external data sources. We're starting with full support for OpenAI Agents workflows, with more frameworks to follow soon.
To all AI builders and security pros — we’re excited to roll out a powerful new upgrade to Agentic Radar that brings even more transparency and visibility into your agentic workflows.
Agentic Radar now supports automatic detection of Model Context Protocol (MCP) server connections, giving teams deeper insight into how agents access external data sources. We're starting with full support for OpenAI Agents workflows, with more frameworks to follow soon.
To all AI builders and security pros — we’re excited to roll out a powerful new upgrade to Agentic Radar that brings even more transparency and visibility into your agentic workflows.
Agentic Radar now supports automatic detection of Model Context Protocol (MCP) server connections, giving teams deeper insight into how agents access external data sources. We're starting with full support for OpenAI Agents workflows, with more frameworks to follow soon.
What is MCP and Why Does It Matter?
Model Context Protocol (MCP) is an emerging open standard that simplifies how AI applications access external data sources – from SaaS tools like Slack and GitHub to internal databases and content platforms.
Instead of hardcoding dozens of brittle APIs, MCP lets developers spin up MCP servers that expose data in a structured and secure way. Agentic apps connect to these servers to retrieve live and context-rich information.
But with that convenience comes a big question:
Do you know which external systems your agents are talking to?
Model Context Protocol (MCP) is an emerging open standard that simplifies how AI applications access external data sources – from SaaS tools like Slack and GitHub to internal databases and content platforms.
Instead of hardcoding dozens of brittle APIs, MCP lets developers spin up MCP servers that expose data in a structured and secure way. Agentic apps connect to these servers to retrieve live and context-rich information.
But with that convenience comes a big question:
Do you know which external systems your agents are talking to?
Model Context Protocol (MCP) is an emerging open standard that simplifies how AI applications access external data sources – from SaaS tools like Slack and GitHub to internal databases and content platforms.
Instead of hardcoding dozens of brittle APIs, MCP lets developers spin up MCP servers that expose data in a structured and secure way. Agentic apps connect to these servers to retrieve live and context-rich information.
But with that convenience comes a big question:
Do you know which external systems your agents are talking to?
Why MCP Detection Is Critical for Security
Since its recent release MCP is gaining traction fast – and for a good reason. But each MCP server connection represents a potential point of exposure. With this new Agentic Radar release, you can now detect:
Data Exposure: Which specific data sources (Slack, GitHub, Google Drive, Postgres, custom databases) are your agents accessing? Understanding this is vital for data governance.
Access Control: Are the permissions granted through the MCP connection appropriate and minimized (least privilege)?
Attack Surface: Each connection point is part of your system's attack surface. Knowing these connections exist is the first step to securing them.
Vulnerability Management: Are the MCP servers themselves, or the underlying data sources, secure?
This is about visibility and control – key pillars of any secure AI deployment.
Since its recent release MCP is gaining traction fast – and for a good reason. But each MCP server connection represents a potential point of exposure. With this new Agentic Radar release, you can now detect:
Data Exposure: Which specific data sources (Slack, GitHub, Google Drive, Postgres, custom databases) are your agents accessing? Understanding this is vital for data governance.
Access Control: Are the permissions granted through the MCP connection appropriate and minimized (least privilege)?
Attack Surface: Each connection point is part of your system's attack surface. Knowing these connections exist is the first step to securing them.
Vulnerability Management: Are the MCP servers themselves, or the underlying data sources, secure?
This is about visibility and control – key pillars of any secure AI deployment.
Since its recent release MCP is gaining traction fast – and for a good reason. But each MCP server connection represents a potential point of exposure. With this new Agentic Radar release, you can now detect:
Data Exposure: Which specific data sources (Slack, GitHub, Google Drive, Postgres, custom databases) are your agents accessing? Understanding this is vital for data governance.
Access Control: Are the permissions granted through the MCP connection appropriate and minimized (least privilege)?
Attack Surface: Each connection point is part of your system's attack surface. Knowing these connections exist is the first step to securing them.
Vulnerability Management: Are the MCP servers themselves, or the underlying data sources, secure?
This is about visibility and control – key pillars of any secure AI deployment.
How It Works
Agentic Radar, our security scanner for agentic workflows, now automatically identifies defined connections to MCP servers within your OpenAI Agents codebase.
Running the scan is straightforward:
Install the tool:
pip install agentic-radarPrepare the input data. Use your own code or copy one of the examples from here.
Run Agentic Radar with:
agentic-radar -i path/to/your/example -o report.html openai-agentsOpen the generated
report.htmlfile in the browser of your choice.
Agentic Radar, our security scanner for agentic workflows, now automatically identifies defined connections to MCP servers within your OpenAI Agents codebase.
Running the scan is straightforward:
Install the tool:
pip install agentic-radarPrepare the input data. Use your own code or copy one of the examples from here.
Run Agentic Radar with:
agentic-radar -i path/to/your/example -o report.html openai-agentsOpen the generated
report.htmlfile in the browser of your choice.
Agentic Radar, our security scanner for agentic workflows, now automatically identifies defined connections to MCP servers within your OpenAI Agents codebase.
Running the scan is straightforward:
Install the tool:
pip install agentic-radarPrepare the input data. Use your own code or copy one of the examples from here.
Run Agentic Radar with:
agentic-radar -i path/to/your/example -o report.html openai-agentsOpen the generated
report.htmlfile in the browser of your choice.
What You’ll See in the Report
In the visualization section, you’ll now see MCP Servers represented alongside Agents and Tools. The diagram highlights:
Which agents connect to which MCP servers
Which tools interact with external systems via MCP
And in the details table below, you’ll get key metadata on each MCP connection: type, initialization params, and more.
In the visualization section, you’ll now see MCP Servers represented alongside Agents and Tools. The diagram highlights:
Which agents connect to which MCP servers
Which tools interact with external systems via MCP
And in the details table below, you’ll get key metadata on each MCP connection: type, initialization params, and more.
In the visualization section, you’ll now see MCP Servers represented alongside Agents and Tools. The diagram highlights:
Which agents connect to which MCP servers
Which tools interact with external systems via MCP
And in the details table below, you’ll get key metadata on each MCP connection: type, initialization params, and more.
What’s Next?
Detecting MCP connections in OpenAI Agents is just the beginning. As adoption of the Model Context Protocol grows and more frameworks integrate support, we’ll continue expanding Agentic Radar’s capabilities across the agentic ecosystem.
Up next, we’re focused on deeper security features:
Automated red teaming of agent-MCP interactions
More advanced vulnerability scanning for exposed data sources
Security scoring and SBOM-style insights for data flows
Want to try it out? Head over to the Agentic Radar GitHub repo, run a scan on your own code, and see what your agents are really connected to.
Got ideas or feedback? Join our Discord community and shape Agentic Radar with us.
Let’s continue to build a future of secure, transparent, and auditable Agentic AI.
Detecting MCP connections in OpenAI Agents is just the beginning. As adoption of the Model Context Protocol grows and more frameworks integrate support, we’ll continue expanding Agentic Radar’s capabilities across the agentic ecosystem.
Up next, we’re focused on deeper security features:
Automated red teaming of agent-MCP interactions
More advanced vulnerability scanning for exposed data sources
Security scoring and SBOM-style insights for data flows
Want to try it out? Head over to the Agentic Radar GitHub repo, run a scan on your own code, and see what your agents are really connected to.
Got ideas or feedback? Join our Discord community and shape Agentic Radar with us.
Let’s continue to build a future of secure, transparent, and auditable Agentic AI.
Detecting MCP connections in OpenAI Agents is just the beginning. As adoption of the Model Context Protocol grows and more frameworks integrate support, we’ll continue expanding Agentic Radar’s capabilities across the agentic ecosystem.
Up next, we’re focused on deeper security features:
Automated red teaming of agent-MCP interactions
More advanced vulnerability scanning for exposed data sources
Security scoring and SBOM-style insights for data flows
Want to try it out? Head over to the Agentic Radar GitHub repo, run a scan on your own code, and see what your agents are really connected to.
Got ideas or feedback? Join our Discord community and shape Agentic Radar with us.
Let’s continue to build a future of secure, transparent, and auditable Agentic AI.
Ready to leverage AI with confidence?
Ready to leverage AI with confidence?
Ready to leverage AI with confidence?
