Daily Work

The Daily Work MCP Server provides a local, extensible way to record, query, and summarize your daily tasks. Built on FastAPI and SQLite, it exposes a set of MCP tools that let you add entries, query by day or week, search text, export data, and view tag-based summaries. It is useful for keeping a consistent work log and easily sharing insights with your IDE or other MCP-enabled clients.

How to use

To start using the server, you can run it locally in one of two modes. In standard local mode, you run the MCP server as a straightforward STDIO process and connect your MCP client to the running session. If you prefer to expose a remote, streamable HTTP interface, you can start the server with a streamable HTTP transport and access the MCP surface at a dedicated HTTP endpoint.

Once the MCP server is running, you can perform as many operations as you need through your MCP client. The available tools include creating new work entries, listing entries by day or week, searching through entries, summarizing recent work by tags, and managing tags. Use the client to call these tools in sessions, and you can script repetitive tasks or integrate them into your existing workflows.

How to install

Prerequisites: you need Python and a package manager. This setup uses Conda to create an isolated Python environment and pip to install dependencies.

conda create -n daily-work-mcp python=3.11 -y
conda activate daily-work-mcp
pip install -r requirements.txt

Initialize the SQLite database before the first run. This sets up the necessary schema for storing entries and tags.

python scripts/init_db.py

Start the HTTP API (optional). This exposes REST endpoints for interacting with entries directly via HTTP.

uvicorn app.main:app --reload

Start the MCP Server. This runs the local MCP server in STDIO mode by default. You can also choose to run it in a streamable HTTP mode to expose a port for HTTP access.

python scripts/run_mcp_server.py

If you want to use streamable HTTP transport, run this instead to expose a port and enable stateless HTTP for MCP calls at the designated endpoint.

python scripts/run_mcp_server.py --transport streamable-http --host 127.0.0.1 --port 8888

Additional notes

When using the HTTP/Streamable mode, the MCP surface is accessible at http://127.0.0.1:8888/mcp. In debugging scenarios you can simulate a client by sending JSON-RPC requests with Accept headers for both JSON and server-sent events.

The MCP server exposes a set of core tools you can call from your MCP client. They are: record_work_entry, list_work_entries, search_work_entries, summarize_recent_work, check_today_recorded, list_tags, create_tag.

Available tools

record_work_entry

Add a new work entry with a description and optional tags to your daily log.

list_work_entries

Query entries by daily or weekly range to review work completed in a period.

search_work_entries

Perform a keyword search across work entries to locate relevant notes.

summarize_recent_work

Generate tag-based summaries for the most recent N days to highlight focus areas.

check_today_recorded

Check whether you have an entry for today to help with reminders.

list_tags

List all existing tags used in entries.

create_tag

Create a new tag for categorizing entries.