Verqury organizes the whole process around the chat AIs, terminal agents, editors, and browsers you already use — durable project memory, reusable guidance, captured artifacts, one-click session bootstrapping, task routing, and a running count of the build time behind it all.
The problem
It moves constantly — chat ideation, PRD and architecture, terminal coding agents, browser agents, external tools. Context scatters across a dozen surfaces, and nothing holds the thread from first concept to release. Verqury is the control layer around that motion.
What it is
Not a chat window, not a code editor, not an agent orchestrator. A workflow layer — deliberately.
Narrative, decisions, and stage tracking — concept to release — as durable files an agent can read natively.
Skills, standards, and project instructions as first-class, composable, reusable markdown assets.
A global hotkey turns copied prompts, commands, and snippets into searchable, reusable objects.
Assemble the right context packet for the right surface — chat, terminal, browser — in one action.
Route work to direct execution, automation, or a human, and feed completion reports back into memory.
Away from your desk? Approve build decisions from your phone — a tap on Telegram, a typed reply, or long-form context by email. The agent raises its hand; you decide.
Run your coding agent beside the project it belongs to. Multiple tabs, each with its own color, a bell that blinks instead of stealing your keyboard, and drag a file in to type its path.
How much work is actually in this project? Verqury reads the transcripts your agent already leaves on disk and answers in active build time — a coffee break isn't coding. Past sessions are backfilled, so the meter opens with real history, not a zero. Lines of code and true cost, optionally.
Every AI tool is a launch command and a handoff template in config. Add a new chat, agent, editor, or browser without touching a line of code.
How it works
Files are the database. All durable data is plain markdown with YAML frontmatter on disk. SQLite is only a deletable, rebuildable search index — so any terminal agent can read and write your project memory natively. Model-agnostic by construction, with zero lock-in.
See it
Philosophy
Markdown + frontmatter on disk; the index is a cache. Your data is yours, git-versionable, agent-readable, forever.
Every feature works from the verqury CLI before it gets a UI. Agents can drive the whole thing.
AI surfaces are launch commands + handoff templates in config. Adding a new tool never needs a code change.
Always available, always organized, never demanding attention. A layer, not a destination.
Get it
Under the hood
Verqury is built in the open as a portfolio of considered engineering — read the reasoning, not just the code.