Agentic Radar CLI
Agentic Radar analyzes multi-agent code on LangGraph, CrewAI, n8n, OpenAI Agents and AutoGen and builds an HTML report on tools and vulnerabilities
- 1,058 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The `test` command runs your project's script with adversarial inputs and sends the check data to OpenAI or Azure OpenAI
- The `--harden-prompts` flag sends agent system prompts to an external service
- Static analysis reads the project sources in full but does not execute the code
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agentic-radarDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The `test` command runs your project's script with adversarial inputs and sends the check data to OpenAI or Azure OpenAI
- The `--harden-prompts` flag sends agent system prompts to an external service
- Static analysis reads the project sources in full but does not execute the code
This is third-party code. Review the repository files before installing.
What it does
Agentic Radar is a Python command line program from SPLX. `agentic-radar scan` reads the sources of a project on the chosen framework (`langgraph`, `crewai`, `n8n`, `openai-agents` or `autogen`) and builds an HTML report without running the code. The report holds a graph of the agent workflow, a list of built-in and custom tools, detected MCP servers and a table that maps each tool to known vulnerabilities from the OWASP Top 10 for LLM applications and OWASP Agentic AI Threats and Mitigations. The `--harden-prompts` flag also sends the detected system prompts to OpenAI, and improved versions following prompt engineering practices come back into the report. It works for OpenAI Agents, CrewAI and AutoGen, and the `OPENAI_API_KEY` key is read from the environment. `agentic-radar test` checks a system at run time. It starts your entry script, feeds it a set of adversarial inputs (prompt injection, personal data leakage, harmful content, fake news) and shows a table of what the agents answered. A language model judges whether an attack succeeded against the condition in the test, and you add your own checks as a YAML file through `--config`. This command supports only OpenAI Agents for now and needs `OPENAI_API_KEY` or `AZURE_OPENAI_API_KEY`.
Who it is for. For developers and security specialists who build multi-agent systems and want to see how they are wired and where they are weak before release.
Good fit when
- A LangGraph, CrewAI, n8n, OpenAI Agents or AutoGen project grew and you need a map of who calls whom and which tools are involved
- You need to check which MCP servers are wired into the project's agents: the report lists them for each supported framework
- You want agents checked on every push: `examples/github_workflow` holds a ready GitHub Actions workflow and the report is saved as a run artifact
- Before releasing an OpenAI Agents agent you want to run it against prompt injection and personal data leakage with `agentic-radar test`
- Agent system prompts need a structured form: `--harden-prompts` shows improved versions in the report
Not a fit when
- Your agents use another framework: only LangGraph, CrewAI, n8n, OpenAI Agents and AutoGen are supported
- You want to vet individual MCP servers or skills before installing them: Agentic Radar reads the code of your own agents and does not connect to third-party servers; mcp-scanner and skillspector do that
- Your agents do not use OpenAI and prompts must not go to external models: `test` needs an OpenAI or Azure OpenAI key and `--harden-prompts` sends prompts to OpenAI
- You need a security guarantee: the report lists possible vulnerabilities by OWASP lists and is not the result of an attack
Example request
Scan the LangGraph agents folder and build an HTML report on the tools and MCP servers they useLimitations
- The package metadata marks the project as alpha (0.14.1), with the last update in November 2025.
- Not every feature works with every framework: runtime testing supports only OpenAI Agents, and prompt hardening supports OpenAI Agents, CrewAI and AutoGen, but not LangGraph or n8n.
- CrewAI needs extra dependencies, `pip install "agentic-radar[crewai]"`, which install on Python 3.10 to 3.12. On other versions the built-in tool descriptions are thinner or missing.
- Static analysis runs locally and does not send code out. Only system prompts under `--harden-prompts` and the data of `test` leave the machine.
- The `test` command injects itself into your process and runs your system. Run it in an isolated environment with a separate key.
How to disable. Agentic Radar runs on demand and does not hook into the project. Remove it with `pip uninstall agentic-radar`. If you added the GitHub Actions workflow from the example, delete its file from `.github/workflows`.
FAQ
Which frameworks can Agentic Radar analyze?
`scan` accepts `langgraph`, `crewai`, `n8n`, `openai-agents` and `autogen`. Each gets a graph, a tool list and MCP server detection. The other features do not work everywhere, and the capability table sits in the roadmap section.
Does my code leave the machine during a scan?
No, static analysis and vulnerability mapping run locally. According to the authors, only data of optional features can go out, for example system prompts when hardened with `--harden-prompts`.
What does the test command check?
It runs your script and feeds it tests for prompt injection, personal data leakage, harmful content and fake news. The result prints as a table with agent name, test type, input, answer and outcome. You add your own tests in a YAML file with `name`, `input` and `success_condition` fields.
Can the check run automatically?
Yes, the repository has an `auto-agentic-radar.yaml` example for GitHub Actions. It runs a scan on every push and saves the report as a run artifact.
How do I turn off the built-in tests and keep only mine?
In the YAML file for `--config` set `include_default_tests: false` and list your tests under `tests`. By default the built-in tests run together with yours.
Discussion
No questions or reviews yet.
Related
CLI toolsAgentic SAST scanner: AI agents read your code, confirm findings through the call graph and scan a whole repository or a diff for pull request review
Open-source adversary emulation for AI agents, LLM apps and MCP servers, checked against OWASP catalogs
AgentShield audits the .claude folder for plaintext keys, broad permissions, dangerous hooks and risky MCP servers and grades the setup from A to F
Burp Suite extension with AI: analyzes requests, runs scanners and opens Burp over MCP to Claude Desktop, Claude Code, Codex and Gemini CLI
An authorized bug-bounty suite: 50 subagents, 26 commands and 2 bug-bounty-platform MCP servers for Claude Code, Codex and five more agents
A CLI with policies that intercepts dangerous AI agent commands like rm -rf, force push or reading .env before they run
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| Agentic Radar (this entry) | yes | 1,058 | |
| MCP Scanner | yes | 1,096 | connects to MCP servers and checks their tool descriptions; Agentic Radar reads the code of your own agents |
| SkillSpector | no | 21,005 | vets third-party skills and MCP servers before install by vulnerability patterns; Agentic Radar analyzes your own multi-agent projects |
README in short
The documentation opens with a description of the report and a link to an example HTML. Next come installation, the `scan` and `test` commands, prompt hardening, custom tests, a CI workflow, a framework support matrix, a list of the developers' articles and an answer on what happens to your code. The repository has a package with analysis modules for each framework, an `examples` folder, a changelog and contribution rules.