Artel is where ideas mature into projects.

Most project tools are flatpacks of features. Confluence has docs, Jira has tasks, Linear has timelines, GitHub has commits. Each is fine; none of them know each other. The seam between them is where context goes to die.

Artel keeps them in one room. Docs, tasks, decisions, and whiteboards live as siblings — and the AI features that ship in AN7 know all of them, because they all share the same project context graph.

The single most expensive thing about working with coding agents today is re-establishing context every time you open a new chat. Artel solves this with an MCP server per project. Connect a project once in Flow Console; from then on, every agent session opens with the project's docs, current tasks, decision log, and architecture notes already loaded.

Agents can also write back: log a decision, update a task, append to a doc, attach a commit. The audit trail is automatic — the project always shows what the agent did, with the reasoning intact.

• AN1 (now) — workspaces, projects, members, MCP discovery stub.
• AN2 — Tiptap docs with a page tree per project.
• AN3 — tasks + timeline (table, kanban, Gantt).
• AN4 — tldraw whiteboards as first-class doc embeds.
• AN5 — Flow Console MCP connector goes live with real tool dispatch.
• AN6 — GitHub App bindings (issues, PRs, commits).
• AN7 — Pro-tier AI features (docs Q&A, decision-log assistant, timeline forecasting).
• AN8 — serendipity polish, calm-by-default UI, domain mapping to artel.metantel.com.

The palette is sage-and-slate, not the manic orange of a Reddit clone. The animations are slow on purpose. The home page surfaces one forgotten doc per day instead of an infinite firehose. Serendipity is a feature, not an accident.