Vaikora › Blog › Product Updates

vaikora-guard-mcp is on PyPI: install it, point it at your agent, get policy decisions before tool calls run

Product Updates · Nov 10, 2026 · 3 min read

If you run an AI agent that calls tools through MCP, right now every tool call executes unless your application code explicitly stops it. There is no enforcement layer between the agent and the tool. vaikora-guard-mcp adds one, and it is available today on PyPI under the MIT licence.

What the package actually does

vaikora-guard-mcp is a self-hosted proxy that sits inline between an MCP client (your agent) and the MCP server (the tool). Before any tool call reaches the tool, the proxy evaluates it against policy and returns one of four decisions: allow, block, modify, or require-approval.

That decision happens before execution. If the policy says block, the tool never runs. No after-the-fact detection, no log entry you review tomorrow morning.

Every decision is written into a SHA-256 hash-chained audit log. Each entry is signed into the chain, so the record is ordered and tamper-evident. That matters when a compliance auditor asks what your agent did at 2:47 AM on a specific date.

Installing it

```
pip install vaikora-guard-mcp
```

Point your MCP client at the proxy instead of directly at the tool server. No SDK migration. The proxy forwards allowed calls and handles blocked or approval-required ones according to your policy configuration.

The source is on GitHub, MIT licensed, written in Python. You can read every line of enforcement logic before you deploy it.

How policy is defined

Policy rules are deterministic: identical input always produces identical output. There is no model scoring a call as "probably fine." The proxy checks the tool name, the arguments, and any configured content inspection rules, then applies the matching policy rule.

Vaikora Guard MCP ships six content inspection modules out of the box. A concrete example: you configure a rule that blocks any tool call where an argument contains a string matching a credit card pattern. An agent that constructs a database write with a PAN in the payload hits that rule. The call is blocked. The agent receives a structured refusal. The audit log records the tool name, the matched rule, the timestamp, and the hash-chain entry.

That is a PCI-DSS relevant control enforced at the tool call layer, not in your application code, and not retrofitted after the fact.

Connecting to the enterprise control plane

If you need centralized policy management across multiple agents or multiple teams, vaikora-guard-mcp connects to the Vaikora Control Plane, hosted by Data443. The control plane is where you manage policy versions, review audit logs across all proxies, and configure SIEM integrations with Microsoft Sentinel, CrowdStrike Falcon, SentinelOne, and AWS Security Hub.

The open-source proxy and the enterprise control plane are separate products. You can run the proxy standalone with local policy files and never touch the control plane. The MIT licence means there is no restriction on how you deploy it.

Why enforcement before execution matters for MCP

MCP tool calls are fast. An agent running autonomously can call dozens of tools in seconds. A detection-after-the-fact approach means the write, the API call, or the data exfiltration already happened. Enforcement before execution is the only architectural position that can actually prevent an action, not just record it.

The proxy adds P50 8ms and P99 45ms to each call at up to 10,000 actions per second, so the overhead does not become a bottleneck in production agent workloads.

Compliance presets for HIPAA, PCI-DSS, GDPR, SOC 2, and ISO 27001 ship with the package. Each preset is a policy file you load. You can extend or override any rule.

What to check next

Install the package from PyPI and read the policy configuration documentation before you write your first rule. The enforcement logic is in the source, so verify that the content inspection modules cover the tool call patterns your agents actually generate. If you are running agents that write to databases, call external APIs, or handle regulated data, that is where to start: map each tool type to a policy rule and confirm the audit log captures what your compliance posture requires.

More from the Vaikora blog