Six content inspection modules. Adapters for the major hosted model APIs. A SHA-256 hash-chained audit log. All of it available on PyPI right now as `vaikora-guard-mcp`, MIT-licensed, self-hosted, no vendor lock-in. If you are running AI agents through MCP and want deterministic policy checks before each tool call executes, that is the short version of what this release gives you.
What the package actually does
Vaikora Guard MCP sits inline between an AI agent and the tools it calls. Before any tool call executes, the gateway evaluates it against your policy and returns one of four decisions: allow, block, modify, or require-approval. The agent never reaches the tool unless the policy check passes. This happens before execution, not after.
That distinction matters operationally. Post-execution detection tells you something went wrong. Pre-execution enforcement means the action either meets policy or it does not run. Identical input always produces an identical decision, which makes the system auditable and the behavior predictable.
Every decision is written into a SHA-256 hash-chained, tamper-evident audit log. If you need to reconstruct what an agent did and why each action was allowed or blocked, the log gives you an ordered record that cannot be silently altered.
Enforcement latency is P50 8ms, P99 45ms, at over 10,000 actions per second. That is the measured overhead of adding a policy gate to every tool call.
What is inside the package
The open-source gateway ships six content inspection modules. These cover the categories of content that matter most when an agent is calling external tools: data exfiltration patterns, prompt injection attempts, sensitive credential exposure, and similar signals. The inspection happens at the gateway layer before the call reaches the downstream tool or LLM.
The package includes adapters for the major hosted model APIs, so you can connect to existing infrastructure without rewriting your agent. It supports 12 LLM providers with no SDK migration required.
`vaikora-guard-mcp` is written in Python and published on PyPI. Installation is:
“`
pip install vaikora-guard-mcp
“`
Because it is self-hosted, the gateway runs in your environment. Your traffic does not transit Data443 infrastructure unless you choose to use the enterprise control plane. The MIT license means you can read, fork, modify, and distribute the code without restriction.
How MCP compatibility changes the deployment picture
MCP (Model Context Protocol) is the interface that lets AI agents call tools in a standardized way. An MCP-compatible proxy can intercept those calls without the agent needing to know a policy layer exists. The agent sends a tool call request. The gateway receives it, evaluates policy, and either forwards the request or blocks it. From the agent’s perspective, the interface does not change.
This means you can add enforcement to an existing MCP-based agent workflow without redesigning the agent. The gateway is the enforcement point. The agent is unchanged.
A concrete scenario: an agent running a customer support workflow is given access to a CRM tool and a file-read tool. Without a policy gate, a prompt injection in a customer message could cause the agent to read files outside its intended scope. With Vaikora Guard MCP inline, a policy rule restricting file-read to specific directories blocks that call before it executes. The audit log records the block with the full context of the attempted call.
Compliance presets and the control plane
The open-source package handles enforcement at the gateway. The enterprise control plane, hosted by Data443, handles policy management, team-level configuration, reporting, and integration with external security tooling.
Compliance presets ship for SOC 2, HIPAA, GDPR, PCI DSS, and ISO 27001. These are starting configurations, not a substitute for reviewing what your specific environment requires, but they reduce the time to get basic policy coverage in place.
Vaikora also integrates with Microsoft Sentinel, Microsoft Defender, SentinelOne, CrowdStrike Falcon, and AWS Security Hub. High-severity agent threats are pushed into SentinelOne as threat intelligence indicators with severity mapping. Critical maps to 100, High to 75, Medium to 50, Low to 25. High-severity risk is also promoted into Azure Security Center and Microsoft Defender for Cloud as native incidents, with no custom engineering required.
For teams already using those platforms, that means AI agent enforcement signals appear in the same workflows used for other threat response, without building a separate pipeline.
What to check next
Start with the PyPI package page for `vaikora-guard-mcp`. Review the six content inspection modules against the categories of risk your agents actually generate. If your agents touch regulated data or call external APIs with real credentials, those modules are the first thing to configure.
Check whether your current MCP server already produces an audit log for tool calls. If it does not, the hash-chained log in the gateway is a straightforward way to get one. If it does, confirm whether that log is tamper-evident and whether the format is compatible with your SIEM.
Vaikora is also available on AWS Marketplace as a SaaS subscription and in the Azure Sentinel content hub as the Vaikora-AzureSecurityCenter solution, if a managed deployment fits your environment better than self-hosted.