Casefile — Claude Code MCP server: what it does, how to install
A local Docker service tracks task progress and decisions across AI agent sessions. It keeps context intact when sessions reset.
Open on GitHubWhat Casefile does
Casefile deploys a local work log in Docker, giving Claude Code and Codex a shared space for task updates. Unlike editable note files, every case entry is append-only. Agents write down design decisions and dead ends as they proceed.
When a session fills its context window, the next agent reads the task card, latest summary, and unresolved questions. It picks up immediately without scanning the entire repository again. A local web board at http://localhost:8080 lets you inspect ongoing work live.
Docker is required to run the service. The installer fails if the Docker engine is offline, because it pulls images and wires an updater container directly through the Docker socket.
What it helps with
- carry task context across agent sessions after a context reset
- track dead ends and abandoned attempts to prevent repeated mistakes
- prevent ticket closure before required checks and summaries pass
- monitor multi-agent progress from a local browser dashboard
Example prompts for the agent
- Create a new task for the database migration and document our initial findings in Casefile.
- Review the open questions on the current task and report the latest verdict.
- Log our failed test attempt and explain why that approach broke the build.
How to install Casefile
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Casefile
How to install Casefile in Claude Code?
Run curl -fsSL https://raw.githubusercontent.com/azimov777/casefile/main/install.sh | sh on macOS or Linux (or irm https://raw.githubusercontent.com/azimov777/casefile/main/install.ps1 | iex on Windows), then sign in with claude mcp login plugin:casefile:casefile. Make sure Docker is running before executing the script.
What does Casefile do?
It maintains an append-only task log tracking decisions and blocked steps so different agents can hand off work cleanly.
How does Casefile differ from memory servers?
It skips semantic vector search and delivers a structured work log in a fixed sequence: task details, latest summary, and open questions.
