tasker
Enables AI agents to manage hierarchical tasks and stories stored as Markdown files, providing tools for creating, listing, editing, and updating task status through the Model Context Protocol.
README
Tasker
A Markdown task tracker designed to live in git alongside your code. Tasks are plain Markdown files inside a .tasker/ directory, with hierarchical stories and subtasks, a CLI for humans, and an MCP server so AI agents can manage the same task list.
Installation
Install with pipx (recommended — installs in an isolated environment):
pipx install mcp-tasker
Or with pip:
pip install mcp-tasker
For development (requires uv):
git clone https://github.com/greendwin/mcp-tasker.git
cd mcp-tasker
uv sync --group dev
Quick Start
# Initialize tasker (or let it auto-detect from any subdirectory)
tasker init
# Or initialize a user-level tasker in your home data directory
tasker init --user
# Create a story
tasker new "Build authentication"
# Add subtasks
tasker add s01 "Design login flow"
tasker add s01 "Implement JWT tokens" --details "Use RS256 signing"
# Work on a task
tasker start s01t01
# View what's on your plate
tasker list
# Mark tasks done
tasker done s01t01
# Edit a task in your editor
tasker edit s01t02
Tasks are stored as Markdown in .tasker/ and committed with your code:
.tasker/
s01-build-authentication/
README.md
s01t01-design-login-flow.md
s01t02-implement-jwt-tokens.md
Usage
Create tasks
tasker new <title> # new root story
tasker new <title> --details "..." --slug <slug> # with description and slug
tasker new <title> --editor # create and open in editor
tasker add <parent-id> <title> # inline subtask
tasker add <parent-id> <title> --details "..." # subtask with description
tasker add <parent-id> <title> --editor # create and open in editor
tasker add-many <parent-id> # add multiple subtasks interactively
Update status
tasker start <task-id>... # mark in-progress
tasker review <task-id>... # submit leaf task for review
tasker done <task-id>... # mark done
tasker cancel <task-id>... # cancel
tasker reset <task-id>... # reset to pending
tasker reset <task-id> --force # force reset non-pending subtasks
# Force-close a parent with open subtasks
tasker done <task-id> --force
# Close every in-review task along with any explicitly listed ones
tasker done --reviewed
View tasks
tasker list # all open root tasks
tasker list -a # include closed tasks
tasker list --todo # only tasks from the TODO list
tasker list --archived # list archived tasks
tasker list --closed # show 5 most recently closed tasks
tasker list <task-id> # subtasks of a specific task
tasker view <task-id> # full task details
TODO list
Pin tasks you're actively focused on. The list lives in .tasker/.todo (git-ignored), and archived tasks are removed automatically.
tasker todo <task-id>... # pin task(s) to the TODO list
tasker untodo <task-id>... # remove from the TODO list
tasker list --todo # show only pinned tasks
Edit tasks
tasker edit <task-id> # open in $EDITOR
tasker edit <task-id> --title "New title"
tasker edit <task-id> --details "New description"
tasker edit <task-id> --slug new-slug
Organize
tasker move <task-id> --parent <new-parent> # reparent
tasker move <task-id> --root # promote to story
tasker move <task-id> --delete # delete a task
tasker move <task-id> --parent <p> --editor # reparent and open in editor
tasker archive <task-id> # archive completed story
tasker archive --closed # archive all closed stories
tasker unarchive <task-id> # restore from archive
Shortcuts
Reference recent tasks without typing full IDs:
| Shortcut | Meaning |
|---|---|
q |
Last referenced task |
q01 |
Subtask 01 of recent |
p |
Parent of recent |
p03 |
Sibling 03 via parent |
tasker view s01t02 # sets recent = s01t02
tasker start q # starts s01t02
tasker view p # views s01 (parent)
tasker done q01 # marks s01t0201 done
MCP Server
tasker can run as a Model Context Protocol server, allowing AI agents to manage your tasks directly.
Configure in Claude Code
claude mcp add tasker -- tasker mcp
Configure per-project (.mcp.json)
{
"mcpServers": {
"tasker": {
"command": "tasker",
"args": ["mcp"]
}
}
}
If running from a checkout:
{
"mcpServers": {
"tasker": {
"command": "uv",
"args": ["run", "tasker", "mcp"]
}
}
}
HTTP transport
For network-accessible clients, start with --port:
tasker mcp --port 8080
Available tools
Once connected, the MCP server exposes:
| Tool | Description |
|---|---|
create_task |
Create a root task or subtask |
list_tasks |
List all root tasks (pass todo=true for only pinned tasks) |
view_tasks |
View detailed info for multiple tasks |
edit_task |
Update a task's title, description, or slug |
start_task |
Mark task in-progress |
review_task |
Mark task in-review (submit for review) |
reset_task |
Reset task to pending |
finish_task |
Mark task done |
cancel_task |
Cancel a task |
Development
uv sync --group dev
# Run all checks (lint + tests)
uv run tox
# Run tests only
uv run tox -e test
# Lint (black, isort, flake8, mypy)
uv run tox -e lint
# Format code
uv run black src tests
uv run isort src tests
Requirements
- Python >= 3.10
Release Notes
1.5.0
tasker initnow creates.tasker/; legacytasker/directories are still recognised at discovery time- Direct task references accept single-digit segments:
s1→s01,s1t1→s01t01, and trailing-slugworks on the padded form list --revfalls back to the active TODO list when nothing is in review (previously fell back to all active root tasks)- Ambiguous-digit error message reworded to mention direct references
- Bug fixes: suppress spurious
Error: <code>output fromtyper.Exiton normal CLI exits
1.4.2
- Single-digit shortcuts pad to two:
q3→q03,p3→p03,ta3→ta03; odd-length runs > 1 are rejected as ambiguous
1.4.1
list --rev/list --in-reviewshows tasks awaiting review (with parent context); falls back to active root tasks when none are in review
1.4.0
t<letter>shortcuts (ta..tz) for active TODO tasks, usable wherever a<task-id>is acceptedlist --todohides finished tasks while active ones remain; printsAll tasks finished!when none are active- TODO list preserves insertion order
- MCP:
task_refarguments accept the full set of CLI shortcuts (q,p,t<letter>) - Build: cached Typer command tree in tests cuts full-suite runtime by ~30%
- Bug fixes:
edit --editorrefreshes the in-memory task tree so follow-up displays reflect new title/slug
1.3.4
- Warning shown when a task directory is missing its
README.md - Exception callstacks hidden by default, shown only with
--debug - Renamed tasks displayed as a single group in
moveoutput - Bug fixes: broken task files now report the offending filename in the error message
1.3.3
- Task preview after
done/cancelwalks up to the first non-closed ancestor, showing the full picture of remaining work - When closing a task completes the entire story, nearby open stories are shown as a "what's next?" hint
1.3.2
- Bug fixes: proper error handling for exceptions raised in dependency-injection callbacks (e.g. missing
tasker/directory)
1.3.1
list --closedflag to explicitly show recently closed tasks (previously shown implicitly at the end oflist)- Bug fixes:
list --todono longer shows non-TODO tasks
1.3.0
in-reviewstatus andtasker reviewcommand for submitting leaf tasksdone --reviewedcloses every currently in-review task in one calltodo/untodocommands to pin tasks, andlist --todo/(todo)marker- Archived tasks are auto-removed from the TODO list
init --usercreates a user-level tasker directory (respectsXDG_DATA_HOME/LOCALAPPDATA)move --editorto open the moved task after reparentingnewandaddaccept unquoted titles (extra words are joined automatically)- Recently closed tasks are shown at the end of
listoutput - Last two digits of subtask IDs are highlighted in bullet lists
- MCP: added
cancel_tasktool,todoparameter onlist_tasks - MCP: status tools return smaller previews on
start/cancel/review/done - Build: migrated from
poetrytouv - Bug fixes:
[text]escaping in task output, stale(q)reference to deleted tasks,taskerdirectory resolution, recently closed task suppressed on narrow terminals
1.2.0
initcommand and automatictasker/directory discovery (walks up to git root)move --deleteoption to delete tasksreset --forceto force-reset non-pending subtasks(q)/(p)recent-task markers shown inviewandeditcommands- Subtask count shown in
viewcommand - Tab autocompletion for task ID arguments
- Slug validation
- Bug fixes: recent task override, auto-unarchive logic
1.1.0
--editor(-e) option onnewandaddcommands to open the task in an editor after creationlist --archivedto browse archived taskslisthighlights the most recently referenced task- Editing an archived task auto-unarchives it
- Task preview shown after
start,reset,done,cancel,move,new,add, andeditcommands - MCP: added
edit_tasktool for updating title, description, and slug - MCP:
view_tasksaccepts multiple task IDs in a single call - MCP: task subtasks grouped by status in response
--versionflag- Bug fixes: editor slug path, directory cleanup on
move, multi-task preview onstart
1.0.0
piprelease
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.