Hamster syncs your briefs, tasks, and plans into your repo as markdown files — in real time. Your coding agent reads them on every action. No copy-paste. No stale specs. Just say what you want to build.
Compatibility and optionality
When your briefs, plans, and tasks live inside a vendor's tool, switching costs are brutal. A better model drops and you're stuck — or you start over.
Hamster is the central repository of work, not the model. Briefs and plans stay in one place. The agent is interchangeable. Use Claude for deep reasoning, Gemini for speed, an open-weight model for sensitive code — or a different model for every task on the board.
New frontier model ships on Tuesday? Point it at your .hamster/ directory and keep going.
MCP
The Hamster MCP server gives your agent 20 tools to mark tasks done, create subtasks, and update status — directly from the coding session. When the agent finishes a task, it updates the plan. No one has to do it manually.
How it works
Install, initialize, and start working. Your agent has full project context from the first prompt.
Install the CLI
Run the install command and authenticate with your Hamster workspace. Under a minute.
Initialize your repo
Run `hamster init`. Creates a .hamster/ directory and starts syncing your project context.
Work normally
Open your coding agent. The context files are already there. Ask it to build something — it knows what that means.
Key capabilities
Briefs, tasks, and plans land in .hamster/ as markdown the moment they change in Studio.
Mark tasks done, create subtasks, update status — all from the coding session.
Context files are local. Your agent reads them even without a network connection.
One server, every tool. Claude Code, Cursor, Gemini CLI, Codex, and more.
The CLI generates structured project maps so your agent understands the full codebase.
You don't pull context in. It's just there. Work the way you already work.
Free to start. No credit card required.