From shadow MCPs to governed agents: Syndio achieves control of AI access
Syndio is a workplace equity software company that helps employers analyze and close pay and opportunity gaps. Its customers trust it with sensitive compensation data, so any AI tool that touches company systems has to go through Security first.
As AI adoption grew, engineers across Syndio were setting up their own MCP integrations faster than Security could review them. Alicia Muzzleman, Director of Security and Compliance, and her team, needed a way to approve AI access without reviewing every connection one by one or slowing down productivity.
At a glance
4 mos
from trial to org-wide AI gateway adoption
12+
connectors live on day one for Engineering
100%
of agent actions through Barndoor logged with user, agent, tool, and MCP server
Every
tool governed at the access control level, independent of user permissions
Alicia Muzzleman
Director of Security and Compliance, Syndio
“Before Barndoor, we had no central way to control what AI tools could connect to, or to see who was connecting what. Engineers were standing up their own MCP integrations.”
The challenge: AI connections not approved by security
Engineers at Syndio were setting up their own MCP integrations and sharing API keys. Security had no central way to see or control what AI tools were connecting to.
“Before Barndoor, we had no central way to control what AI tools could connect to, or to see who was connecting what,” said Muzzleman. “Engineers were standing up their own MCP integrations.”
At one point, an engineer asked for an open-source MCP token to be dropped into a shared password vault, and Security had to step in. API keys were also being shared with individual engineers. Without a control layer, every new connection would have required a new vendor review.
Eventually, Syndio paused software review tickets for MCP connections until it could put MCP governance in place. The effects reached every team.
The solution: One approved path for AI adoption
Syndio met with several shadow-AI-prevention vendors and ran proof-of-concepts. By March the plan was in shape: a separate tool for shadow AI discovery, and Barndoor for MCP and agent governance.
“Our own evaluation provided us a pathway via Barndoor as the only approved path for connecting AI tools to applications,” said Muzzleman.
The trial started in March across Security and Engineering, covering hosted, official, and custom MCPs. Barndoor was approved and Security launched it to Engineering with SSO login, a user guide, and 12 connectors. Within weeks, it was rolled out to the rest of the company.
Outcomes: Visibility, speed, and wider access
Full visibility
Faster engineering
Stronger security
Wider access for the whole company
“Barndoor opened up a wider set of what the org can use across AI platforms, and put it in one central place,” said Muzzleman. “That one pane has raised efficiency significantly because people spend less time hunting for the right tool or waiting on access, and they work with the same governed set of connections.”
An unexpected benefit: A foundation for custom agents
Syndio didn’t plan for Barndoor to become its foundation for internal agent development.
A working partnership
Syndio’s rollout surfaced new needs along the way, and the two teams worked through them.
It’s been a great partnership so far of identifying more unlocks for our organization and passing them off to the Barndoor team to feature-ship efficiently,” she said.
“We have one place to approve, restrict, and audit AI access to tools, and it was the only workable path for MCP connections,” said Muzzleman.
What’s next: Barndoor Bridge and DLP across every endpoint
Syndio is now installing Barndoor Bridge on all endpoints so it can begin testing DLP rollouts within teams. That will bring local MCPs and desktop usage under the same governance as the rest of the company.
← All customers