A local, read-only panel that shows what your agent is actually doing.
Drag to orbit · the screen is a real capture
Live demo
A genuine recording of the dashboard on a sample plan.
64 combinations of the same honest state.
ETag and 304.
Available at /legacy, forever.
The problem
You either interrupt the agent or tail a log.
Task Dashboard gives the run a surface instead.
An engineering estimate, not an API invoice.
What you get
Three properties are load-bearing.
ETag/304 conditional polling.
Pure standard library.
Exact 127.0.0.1 origins.
Never writes a plan.
A single MCP server.
Pending, Active, Review, Blocked, Complete.
Quickstart
Python 3.11 or newer.
git clone https://github.com/skarL007/task-dashboard.git cd task-dashboard
python -m pip install .
task-dashboard serve \ --root /path/to/project
It prints one machine-readable line.
Connect your agent
Install the optional extra, then register the server.
claude mcp add -s user task-dashboard -- \ <python> -m task_dashboard mcp serve \ --url http://127.0.0.1:55219/
codex mcp add task-dashboard -- \ <python> -m task_dashboard mcp serve \ --url http://127.0.0.1:55219/
agy mcp add task-dashboard <python> -- \ -m task_dashboard mcp serve \ --url http://127.0.0.1:55219/
Three parameterless, read-only tools.
Security
Each is covered by a test.
Full threat model in SECURITY.md.
Straight talk
Stated here rather than discovered later.
Open source
Permissive licence and a documented contract.
LICENSECONTRIBUTING.mdSECURITY.mdCODE_OF_CONDUCT.mdCHANGELOG.md.github/Continuous integration