Giving users' own AI agents a secure way into Assuredly

Assuredly helps organisations complete security assessments and provide the evidence behind every answer.

Completing an assessment usually means searching through policies, documents and systems, then copying the right details into each question by hand. Many users already work with AI tools they trust, but those tools had no secure way to connect to Assuredly.

Assuredly needed to let approved users bring their own AI agents into the platform, without weakening the security its customers rely on.

Upgrowth built an MCP server that lets approved users' agents connect securely to Assuredly and complete assessments on their behalf. We also developed agent skills, plugins and integration kits for ChatGPT Work, Claude Code and Microsoft 365 Copilot, so trial participants could start using it in the tools they already know.‍

Services

Product Strategy
Solution Architecture
Security Architecture
Engineering
AI Integration
Agent Skills and Plugins

Website

ASSUREDLY.CO

What we did

A secure connection built around the tools people already use

We designed the MCP server and its integrations so agents could do useful work inside Assuredly, while users and administrators keep full visibility of what happens and who did it.

Starting with security

Agents can only reach a subset of the features and data their user already has access to. We added security gates at every stage, so each connection is freshly authenticated and every action is auditable and attributable.

A minimal extension to the platform

We added the server as a small extension to Assuredly's existing infrastructure, so the new capability caused as little disruption as possible to the systems customers already depend on.

Enhancing the user experience with agent skills, plugins and integrations

We developed agent skills, plugins and integration kits for ChatGPT Work, Claude Code and Microsoft 365 Copilot, so trial participants could start using it in the tools they already know. Building and testing in the same client software as users helped make the experience as smooth as possible.

Outcome

A faster assessment workflow, with users in control

ChatGPT and Claude Code users install a single plugin and have everything they need in a few clicks. Copilot administrators can use the kit to set up customised agents for a whole team in one action.

Users point the client they already trust at the files and sources that hold their security details. The agent answers the questions, uploads files as evidence and links each answer to the files it was based on. When it can't answer a question, it leaves a note to help the user complete it.

Every AI-provided answer and note is clearly flagged, and users can jump straight to new answers to review them. The result is a much faster and more convenient way to complete assessments, and new value across Assuredly's product offering.