Cycode says its new Citizen Developer Security service will help organisations find and secure AI-assisted software building beyond engineering teams. Available now in private preview, it combines visibility into AI builders and unmanaged codebases with protections for devices, prompts and the applications agents produce.
Watch Desk analysis
What happened
The service is an extension of Cycode’s Agentic Development Security platform. The company says its inventory can identify builders, unmanaged codebases, coding agents, Model Context Protocol servers and other AI tools.
Cycode describes three areas of protection: blocking malicious packages on workstations; applying rules to prompts, agent file reads and MCP calls to help keep secrets and sensitive data out; and scanning generated code, dependencies, secrets and cloud configuration before applications reach production. Its announcement also says the service can block a package that fails a cooldown policy and select an alternative, or generate fixes for vulnerable code.
Why it matters
AI coding tools mean software can be built outside familiar repositories and development pipelines. Cycode’s launch targets that gap, putting security controls around employees who may not be professional developers as well as established engineering teams.
The practical pitch is to find those tools and outputs, then apply company policies without stopping the work altogether. The offering is in private preview, however, and the announcement does not provide independent performance results or evidence of how well the controls work in practice.
Our read
This is a concrete attempt to extend software security into the less orderly corners of AI-assisted development. The most interesting promise is not simply spotting risky code, but catching problems during the build and offering a fix or safer alternative.
That could make security less of a final gate and more of a working guardrail. For now, it is a company announcement about a private-preview product, not proof that the approach prevents vulnerabilities or keeps every sensitive prompt safe.
What to watch
- When the private preview becomes generally available.
- Which AI tools, agents and MCP servers the inventory supports.
- How the controls behave when they block an action or generate a fix.
- Whether Cycode publishes customer results or technical evaluations.
Discussion spark: Should security tools automatically block risky AI-generated code and choose alternatives, or should employees always make that call themselves?
Sources and evidence
- Source update (4 November 2026, 14:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.