Standup Journal MCP

A personal standup journal as an MCP server — log tasks, generate standup reports, and manage a checklist dashboard, all from Claude Desktop.

One Dataset, Two Views

The standup log tools and the checklist dashboard both read from and write to the same underlying task table — they are not separate stores. An item logged with log_task shows up in get_checklist and the dashboard, and vice versa, using the same task IDs.

• Standup log tools (log_task, get_tasks_by_date, get_tasks_between, generate_standup_report) present tasks as a dated journal — good for "what did I do on Aug 12" or generating a Slack-ready standup message.

• Checklist dashboard (get_checklist, add_checklist_item, toggle_checklist_item, delete_checklist_item, open_checklist_dashboard) presents the same tasks as a checkable to-do list at http://localhost:9249.

In short: it's one list of tasks, viewable either as a standup journal or as a checklist — pick whichever framing suits what you're trying to do in the moment.

Related MCP server: work-tracker-mcp

Following Along (YouTube Tutorial)

This project follows MCP Tutorial: Build Your First MCP Server, adapted for uv and the current v2 SDK.

Step 1: Initialize the project

uv init
uv add "mcp[cli]"

Step 2: Write the server

Create main.py in the project root. The server uses MCPServer from mcp.server with @mcp.tool() decorators.

Note: The video uses FastMCP from mcp.server.fastmcp (v1.x SDK). uv add "mcp[cli]" today installs v2, where FastMCP was renamed MCPServer and moved to mcp.server. Decorators are unchanged.

Data lives in ~/.standup-journal/standups.db (override with STANDUP_DB_PATH).

Step 3: Install into Claude Desktop

uv run mcp install main.py

Fully quit and reopen Claude Desktop for the server to appear.

Step 4: Verify in Claude Desktop

Open Settings → Developer → Edit Config to confirm standup-journal is listed under mcpServers. Restart Claude Desktop, then check Connectors — your tools should be available.

Connecting from a Kiro / KiroCrew scheduled job

A Kiro / KiroCrew agent can use this same MCP server the same way Claude Desktop does — you call the tools by asking in chat. It connects exactly the same way as Claude Desktop: the agent spawns the server over stdio using an identical mcpServers entry. On top of that, you can optionally schedule a KiroCrew cron job to wake the agent every weekday morning and auto-generate your standup — but that scheduling is an extra setup step, not something the server does by itself.

Step 1: Register the server on your Kiro agent

Add the same entry Claude Desktop uses to your agent config (~/.kiro/agents/<agent>.json) under mcpServers:

"standup-journal": {
  "command": "/opt/homebrew/bin/uv",
  "args": [
    "run", "--frozen",
    "--with", "mcp[cli]==2.0.0",
    "mcp", "run",
    "/Users/you/standup-journal-mcp/main.py"
  ]
}

Then allow the agent to call the tools by adding "@standup-journal" to the agent's tools and allowedTools arrays, and a permissions rule:

{ "capability": "mcp", "match": ["standup-journal/*"], "effect": "allow" }

This is byte-for-byte the same launch command as the Claude Desktop connector — the only difference is whether Claude Desktop or a Kiro Crew agent spawns it. Because the data lives in ~/.standup-journal/standups.db, Kiro Crew and Claude Desktop see the same task list.

Step 2: Schedule the standup

Create a recurring job that wakes the agent to produce your standup. From the KiroCrew dashboard or CLI:

kirocrew cron add "Daily Standup Report" \
  "Generate my standup report using the standup-journal generate_standup_report tool and send it to me." \
  --agent kirocrew \
  --cron "0 9 * * 1-5" \
  --approval-mode auto

Cron expressions are evaluated in your machine's local timezone, so 0 9 * * 1-5 fires weekdays at 9:00 AM local. --approval-mode auto lets the unattended job call the tools without a prompt. Manage it later with kirocrew cron list, kirocrew cron trigger <id> (run once now), or kirocrew cron pause <id>.

The job wakes the agent on schedule, the agent connects to the standup-journal MCP server the same way Claude Desktop does, calls generate_standup_report, and delivers the result to you.

Note: a Kiro scheduled job is a different mechanism from an MCP connector — the job is what wakes the agent on a timer, and the mcpServers registration is what lets that agent reach these tools. You need both: the registration (Step 1) makes the connection "the same way Claude Desktop does," and the cron (Step 2) is what points a schedule at it.

Tools

Standup log — what you did / are doing / are blocked on

Tags get a consistent royal-themed emoji (👑💎🦋🕯️ etc.) assigned automatically per tag name.

Checklist dashboard — what you need to do

Due dates are parsed from natural language by Claude (e.g. "due Aug 20"). In the dashboard, a due date only appears next to a task once one is set — undated tasks show nothing. Click the 📅 icon on any task to set or edit its due date directly. Tasks with due dates sort to the top; overdue dates show in red.

Interactive Checklist Dashboard

The server starts a local HTTP server on port 9249 (override with CHECKLIST_PORT) alongside the MCP stdio transport. The dashboard at http://localhost:9249 provides an interactive checklist with a progress bar, nested subtasks, due dates, and tag badges. Use it as your working to-do list — items you still need to do, not a record of what's already happened.

The progress bar reflects all-time completion (done tasks ÷ total tasks), not just a recent window — so finishing an older item always moves it. It's a plain sliding bar by design, with no numeric done/in-progress/blocked counts shown alongside it.

# Run standalone (starts MCP server + dashboard)
uv run python main.py

# Custom port
CHECKLIST_PORT=8080 uv run python main.py

Only binds to localhost — not exposed to the network.

Testing in Claude Desktop

Try these:

1. "Log that I finished the login screen styling."

2. "Log that I'm blocked on AWS permissions, tag it infra."

3. "Log that the API docs are due Friday, tag it github."

4. "What have I logged for today?"

5. "Generate my standup report."

Verify data directly:

sqlite3 ~/.standup-journal/standups.db "SELECT * FROM tasks;"

Troubleshooting

• Connector won't connect: Check ~/Library/Logs/Claude/mcp-server-standup-journal.log (macOS) or %APPDATA%\Claude\logs\ (Windows).

• Stray stdout corrupts the connection — use logging (stderr), never print().

Sources: MCP docs, Claude MCP troubleshooting, PyPI MCP

License

Released under the MIT License — free to use, modify, and distribute with attribution and no warranty.