The problem
Project-management tools are built for humans. But when a large part of planning and execution happens with AI agents, copying tasks between a chat window and an app wastes time and focus. I wanted the agent to work directly inside my project manager.
What is the Model Context Protocol (MCP)?
MCP is an open protocol that lets language models talk to external tools and data in a standard way. An MCP server exposes a set of "tools" with well-defined inputs and outputs to the agent.
Projehban's data model
- A project contains phases
- A phase contains tasks, and a task can have subtasks
- Each project's progress is computed automatically from its tasks
- A daily planner places tasks on the calendar in the user's time zone
- An ideas inbox holds work whose time has not come yet
The MCP tools
Projehban's MCP server offers tools such as list projects, create project, add phase, add task, update task, get day and schedule task. I can tell an agent "create the tasks for phase two of the website and plan tomorrow" and see the result immediately in the Android app.
Lessons learned
- Tool naming matters. Agents decide based on tool descriptions; vague descriptions mean vague behaviour.
- Validate inputs strictly. Agents can send bad data too.
- Keep the user's language. Titles are stored in the user's language (Persian) so the app stays natural.
What's next
Projehban is published on GitHub and is being prepared for release on the mobile app stores.