>>> AGENTIC SPEC-SYNCED DEVELOPMENT <<<

The spec is the source of truth.

Living Spec gives engineers and product teams the context they need throughout development. The spec becomes a shared context layer for agentic development.

⚽ At-home Indoor Soccer Arena · Spec

Goal: a room a kid can actually play in. Links: Floor and turf Order turf Drawing: arena-layout

Decision (rev 38): skip the pressure adhesive. Tape between the turf and foam blocks held fine.

rev 39 · agent updated 2m ago · branch merged · acceptance criteria 3/3 ✓ · spec still true

A living spec in the app: the page tree, a page with linked chips and a drawing, and a task showing its recorded decision and merged agent branch

// one document: pages, chips, tasks, drawings and an agent branch, all in place

>>> THE DRIFT <<<

Your code moved. Your spec didn't.

AI and agents now write code faster than any team can capture why. That intent lives in the spec, and the moment it drifts from what you are actually building, everyone is guessing again.

DOCS ROT

Specs end up spread across docs, tickets and READMEs that nobody quite trusts. Two sprints later they describe software that no longer exists.

AGENTS FLY BLIND

Coding agents are only as good as their context. Without the spec they redo yesterday's decisions, badly.

DECISIONS SCATTER

The why ends up in chat threads, meeting notes and somebody's head. By the time you need it, nobody can find it.

/// HOW IT WORKS

One document. Kept true.

Write the spec where work happens

Pages, tasks, drawings and files in one living document. Structured enough for a team, readable by an agent.

>>> FOR YOUR AGENTS <<<

Your agents read the spec too.

Connect Claude, or any MCP client, and your agents read the relevant spec as context. They work from the same source of truth your team does and write their progress back on a branch you review, so the spec stays current as they work. Nobody has to stop and write it up by hand.

GetSpecsSearchPagesCreatePageUpsertSectionCreateTaskEnsureBranchReconcileBranchMergeBranch

// reading this as an agent? start at /llms.txt

>>> THE TOOLBOX <<<

The rest of the toolbox

TASKS

Assign, prioritize, and track work inside the spec.

DRAWINGS

Sketch flows and diagrams that live with the words.

FILES & MEDIA

Attach designs, screenshots, and video inline.

SNIPPETS

Reuse boilerplate blocks across documents.

SEARCH

Find anything across every spec, task, and comment.

SHARING

Invite collaborators, reviewers, or the whole internet.

NOTIFICATIONS

Follow what matters. Ignore the rest.

DASHBOARDS

Roll tasks and progress up across documents.

IMPORT & EXPORT

Markdown, email, PDF and MCP, in and out. No lock-in.

CONTEXT DIAGRAM

A live map of every page, task, and link in the spec.

DYNAMIC PAGES

Saved searches and filters that render as pages, always current.

CLONING

Duplicate any spec, or start from example templates.

INSIGHTS

Sidebars show you connections and activity while you work.

ACCOUNT FUSION

Multiple accounts and email aliases, one unified profile.

FEEDBACK & COMMENTS

Reviews and comments attached to the exact thing they are about.

INTEGRATING...

Works with what you already use

CLAUDE / MCPGITHUBLINEARJIRASLACKASANAEMAILBASECAMPCLICKUP
>>> REV 1 OF YOUR NEXT SPEC <<<

Put your spec to work.

Bring what you already have and connect your agents. The spec stays current from there.

Create your first spec