36 tools, 40 commands: an MCP server and CLI in three months
Gautier Le Poher
Evaboot, June 2026 to August 2026
Situation: the product had to open up to AI assistants.
The core of the app had just been rebuilt on Django, with a public API. Several layers that users and agents use every day were still missing.
The product needed:
a way for AI assistants like Claude or ChatGPT to use it
direct access from the terminal for power users
a conversational way to use it, to ease adoption (through an AI agent)
Mission: ship these interfaces.
These three interfaces had to be built and shipped on the new stack.
Actions
An MCP server. The MCP server is the connector that lets an AI assistant use the product. This one exposes 36 tools: read data, write it, run long tasks and extractions.
A CLI. A tool with 40 commands to use the product from the terminal.
An AI agent. It answers questions on the product's data. The text appears as it is written (streaming), and the agent handles rate limits.
Contributions to the core. I also opened pull requests on the core of the app to add or fix API endpoints.
Results
MCP server in production in August 2026.
CLI at version 0.3.1, installable with a single command.
AI agent tools in production.
Like this project
Posted Oct 20, 2026
I added new interfaces to an existing app: an MCP server (for AI assistants), a CLI (for access from the terminal) and an AI agent for conversational use.
Likes
0
Views
0
Timeline
Jun 1, 2026 - Aug 31, 2026
Clients
Evaboot
