A companion, not a replacement: plnk-tui

plnk-tui is plnk’s TUI companion — a live (Terminal UI) explorer for Planka, with real-time websocket sync.

I didn’t set out to build a Planka client. Planka already has a very polished, responsive web UI that I open and use every day to manage my work.

But as I’ve been working more in partnership with coding agents, I find myself in the terminal more day-to-day — and I’ve been adopting all sorts of TUIs and CLIs to avoid context-switching out to a GUI.

I had no idea any of this existed. TUIs — Terminal User Interfaces — feel like a whole subculture I’d somehow walked past for years. Active communities, mature frameworks in Python and Rust, a long tail of polished tools like LazyGit, LazyDocker, k9s. I lost myself in them for days.

I was building plnk and using it on my own work. Eating my own dog food. For weeks the loop was the same. I’d go look at the Planka web UI, navigate down through project, board, and list, find the cards I wanted the agent to handle, and give it commands like:

On project ABC, find the Work Board, and in the Backlog list let’s find the card with the title that starts with “assess the byz*”.

On the Work Board, move “Refactor session token storage” from In Progress to Review.

Find the cards labeled AuthSystem in the Backlog on the Work Board, scan them and prepare for questions.

Then the agent would finish a card. I’d switch back to the browser, scroll the board, pick the next card, parse it myself, switch back, type in the next prompt. Then again. And again.

My work on plnk had already built an efficient command surface between my agent — running in the terminal — and my Planka instance: a thin, strict CLI that turned Planka’s REST API into a command language the agent could actually use. (More on the why in the previous post.)

But while the agent had a clean working surface and I had the beautiful web UI, the agent and I were missing a common bridge — a shared dataset we could both speak in. I got a bit tired of writing so verbosely to the agent.

I want to be able to speak in Unique Identifiers / IDs, get to the point, cut to the chase.

So I built plnk-tui — a terminal UI that gives me the same hierarchy the CLI already speaks:

project → board → list → card → task

Not a replacement for the Planka web UI. A companion to it, and a bridge — somewhere the agent and I can speak a common language without me jumping out to the GUI every time I need to read or edit a card.

My goal was simple: navigate the Planka hierarchy with my keyboard and, with one keypress — y (YANK) — copy the current context as a JSON string the agent can drop straight into the conversation. One fell swoop, and the agent has exactly where I am. No more describing a card around the agent until it figures out which one I meant.

Here’s what a yank looks like — the full hierarchy, project down to card:

{"project":{"id":"1766510032108651830","name":"BookBridge"},"board":{"id":"1766510291450856764","name":"Work"},"list":{"id":"1766512261381227840","name":"In Progress"},"card":{"id":"1766513265430496590","name":"BB-000: PRD and Design Document"}}

Or, if I’m at the board level and want to talk about every card on it:

{"project":{"id":"1766510032108651830","name":"BookBridge"},"board":{"id":"1766510291450856764","name":"Work"}}

I also provided a plnk command builder in case I just want to run the command myself:

plnk card snapshot 1766513265430496590 --output json

y gives the agent the IDs to use with plnk, Y gives me and the agent an immediate command to run.

Now it’s just a few keypresses — navigate, yank, paste into the agent, and we’re rolling. Less time searching. More time working.

This started out as a read-only hierarchy viewer of Planka, but as time went on I added web-socket listening for live updates — that was an interesting challenge. Now I don’t have to refresh the hierarchy during an agent session that’s actively manipulating it.

Then I added some basic edit capability.

Maybe I’ll add card creation and richer editing later. For now, it’s already doing what I needed — a BRIDGE between my Planka instance and my agent, where we can speak the same hierarchy without me tab-flipping out to the GUI. Less searching. More working. That’s enough for now.