Chat Server
An AI-native workspace built on LLMs and the A2UI protocol, letting teams run daily project activities entirely through conversation.
A2UI Protocol/MCP/LLM/React
The challenge
- Teams lost time and context switching between multiple applications for project management, documentation, and observability, each with its own login and UI.
- Generating reports or querying project data required specific query syntax and dedicated dashboards, creating a learning curve for non-technical users.
- There was no single place to check ongoing work, follow up on tasks, or monitor application health without leaving the conversation and opening separate tools.
What made it hard
- Rendering third-party tool interfaces directly inside a conversational panel required a protocol (A2UI) and an MCP server wired into the main panel itself.
- Supporting plain-English report generation and data queries meant abstracting away query syntax without losing accuracy or flexibility.
- The architecture needed to stay extensible so new connectors could be added as the workspace grew, without redesigning the core channel model.
The approach
- Built Chat Server so any channel starts empty and gains capability through connectors — Backlog (built on TrackIT, an in-house Jira alternative), Canvas and Markdown for wiki-style documentation, and Watchdog for application monitoring and telemetry.
- Implemented an A2UI-native main panel with the MCP server hooked in directly, so tool interfaces render and respond inline inside the conversation rather than redirecting to another screen.
- Enabled plain-English operations for reports, data queries, task creation, and troubleshooting, removing the need to log into or learn each underlying application's UI.
The impact
Replaced logins across multiple applications with a single conversational channel, cut report generation down to plain-English queries, and unified project management, documentation, and monitoring into one workspace.
Next case study
Lab Insights