daily-digest
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@daily-digestShow me today's daily digest and any deadline tasks"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Agent Daily Digest
A local dashboard and CLI that turn files, tasks and deadlines produced by multiple AI assistants into a daily briefing. No model calls, cloud accounts or external network requests are used at runtime.
Install and quick start
Python 3.11+ is required. From a checkout, these three shell commands install the application, prepare fictional examples and open the dashboard:
python3 -m venv .venv && . .venv/bin/activate && pip install -e '.[dev]'
python examples/prepare.py && digest --config examples/config.yaml scan
digest --config examples/config.yaml serveOpen http://127.0.0.1:8761. Use the current date to view six fictional outputs from three assistants. Date selection, assistant filters, previews, task completion, workspace scanning and Markdown export work against the local SQLite database. For a headless check, run digest --config examples/config.yaml today --json instead of the final command.
Runtime-only installation: pip install -r requirements.txt && pip install -e . --no-deps. The example generator needs the dev extra because it creates a synthetic PDF.
Related MCP server: helm-personal-os
Features
Incremental, repeatable indexing of local roots, configurable glob inclusion/exclusion and assistant assignment by glob, prefix or regular expression. Changed assignment rules take effect on the next scan.
Output timeline sorted newest first. Markdown, CSV (first 50 rows), raster images and the first page of a PDF can be previewed.
Deterministic extraction of titles and 200-character summaries, without a language model.
Tasks imported from JSON/CSV, local task commands and an optional read-only connection to
agent-task-board.Overdue / today / this week / later / undated deadline groups; countdowns and daily or ISO-week Markdown reports.
JSON/CSV round-trip exports, dated SQLite online backups and official FastMCP stdio tools.
Configuration and data
See examples/config.yaml. YAML and TOML are supported. Relative configured paths resolve against the configuration file's directory. Without a configuration file, the root is ./workspace. Runtime data defaults to ./data; override it with --data-dir or DIGEST_DATA_DIR. The database is digest.db, generated reports are under reports/ inside the data directory, and backup ZIPs are under backups/.
roots: [workspace]
timezone: Asia/Shanghai
max_file_size: 10000000
include: ['**/*']
exclude: ['**/.git/**', '**/.env', tasks.json, tasks.csv]
agents:
- {glob: 'report/**', name: Report assistant}
- {prefix: 'study/', name: Study assistant}
- {regex: '^content/', name: Content assistant}
countdowns:
- {name: Project review, date: '2026-10-22'}
# Optional: task_sources: [tasks.json, tasks.csv]
# Optional: task_board_db: ../agent-task-board/data/tasks.dbThe optional board adapter reads the tasks table's id, title, assignee, due_at, status, priority and updated_at fields. claimed/running map to doing; review/failed map to blocked. Complete those tasks in the task-board application itself.
Task JSON is an array of objects (or { "tasks": [...] }); CSV uses the same columns:
[{"id":"review-1","title":"Review sample report","agent":"Report assistant","due_at":"2026-10-15","status":"todo","priority":2}]Dates and timestamps use ISO 8601. Date-only deadlines remain due for the entire local calendar day. A bare local timestamp uses the configured timezone. Task imports are deduplicated by source and ID (or a deterministic key); completing an imported task creates a local override retained on subsequent imports. Removing a row from an import file does not delete its existing task. File removal, root removal and exclusion changes remove obsolete file-index entries. Temporarily unavailable roots preserve their previous index and produce a warning.
CLI reference
All commands support --data-dir, --config, --json, --quiet, --version; --progress writes scan progress to stderr. Global options may precede or follow subcommands. No command requires a TTY.
Command | Purpose |
| Index changed files and import tasks |
| Read a daily briefing |
| Filter tasks |
| Add a repeat-safe task |
| Complete a local task |
| Read a file preview |
| Write a daily Markdown report |
| Write an ISO-week report |
| Export all indexed tables (UTF-8 BOM by default) |
| Export JSON tables |
| Merge an export, or import a plain task file |
| Write a dated ZIP with a consistent database snapshot |
| Start the local dashboard |
| Start the stdio MCP server |
JSON responses are a single object: {"ok":true,"data":...,"warnings":[]}. Errors use {"ok":false,"error":{"code":"NOT_FOUND","message":"Task not found"}}. Exit codes: 0 success, 1 operation failure, 2 invalid arguments/data, 3 missing resource. Export CSV uses table,row_json to preserve nested values and types; task-import CSV uses ordinary task columns. Add --no-bom to CSV export for plain UTF-8. Imports merge by primary key; make a backup before importing over existing data.
MCP
Install the package, then configure an MCP client with:
{"mcpServers":{"daily-digest":{"command":"digest","args":["--data-dir","/workspace/data/agent-daily-digest","--config","/workspace/apps/agent-daily-digest/config.yaml","mcp"]}}}Tools: get_digest(date?), list_outputs(date, agent?), list_tasks(status?, due_before?), add_task(title, agent?, due_at?, priority?), complete_task(id). Write tools describe their side effects and share the CLI core. Run digest scan before querying to refresh the index. AGENTS.md contains installation commands, JSON examples and operational recipes.
Verification
ruff check .
ruff format --check .
pytest -q --cov=agent_daily_digestCI runs on Ubuntu with Python 3.11 and 3.12. Tests include fixed-time deadline grouping, deterministic reports, config reassignment, duplicate imports, changed/deleted files, previews, directory escape rejection, transport errors, MCP tools and backups.
FAQ and limitations
Does it watch remote workspaces? No. It scans local configured directories on command; use cron to refresh them.
Is this a multi-user server? No. It defaults to loopback, has no accounts, and rejects cross-origin browser requests. Keep it on a trusted local machine.
Are previews full document editors? No. CSV previews are capped at 50 rows and PDF previews use only first-page text. Scanned PDFs need an external OCR tool.
What does a daily report mean? File dates use modification timestamps in the configured timezone. Rewriting a file moves it to the new date; the index is not an immutable historical file archive. Task completion counts use current indexed state, not a reconstructed historical task state.
How do I restore a backup? Stop the app, preserve the existing data directory, extract
digest.dbinto a new data directory and restore your YAML/TOML configuration from the included JSON representation. Never replace a live WAL database in place.Are private files published? Runtime data, databases, reports and
.envare ignored by Git. Review configured roots yourself before sharing generated reports, which contain local paths and source excerpts.
Roadmap
Optional file-system watch mode, richer historical snapshots and additional task-board adapters.
License
MIT. All bundled example content is fictional.
This server cannot be deployed
Maintenance
Related MCP Connectors
Local-first task manager: create, edit, and complete tasks, projects, and checklists via MCP.
Create, list, and complete todo items through MCP.
Private-by-default, local-first memory/context/task orchestrator for MCP apps and agents.
Manage tasks, Focus Zone, notes, projects, and task history from compatible AI assistants.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceExposes task management (add, list, complete tasks) and document search (RAG) as MCP tools for AI agents.-
- AlicenseNot gradedqualityBmaintenanceProvides MCP tools to read and update a local-first personal OS for goals, tasks, habits, food, workouts, and check-ins. Enables assistants to manage daily life data and interact with an evidence-grounded AI coach.MIT
- AlicenseNot gradedqualityBmaintenanceEnables agents to manage personal task lists through the MCP protocol, supporting task creation, editing, completion, deletion, grouping, reordering, and JSON import/export.1AGPL 3.0
- FlicenseNot gradedqualityBmaintenanceEnables personal productivity management by exposing MCP tools to create, retrieve, list, update, complete, and delete tasks and calendar events, with validation and conflict detection.-