Skip to content

A tour of the app

A project in Minga has five tabs. Each one answers one question, and none repeats another: the Overview talks about the work, Team talks about who does it. This page walks through them with Dictly, an example dictation app.

One row at the top of every project page:

  • the workspace and project switchers;
  • the five tabs: Overview · Chat · Plan · Team · Needs you, with the Needs you count next to its tab;
  • Lea’s status chip, which opens Chat. It reads working, free, queued, away (her runner is offline), paused or can’t work (she needs a model key, or the person who looks after her left the workspace);
  • “1 running” or “1 waiting” when workers are live, which opens the list of runs;
  • project settings, and your avatar, which opens your Account menu.

On a phone the five tabs sit in the tab bar at the bottom, and ⋯ holds runs, settings and your account.

The Overview is the default tab. Every line opens the record behind it; from here you can also add a goal, edit the brief and give a milestone the go-ahead.

  • Brief: the first lines of the vision and of what you are not doing.
  • Milestones: each open milestone with its due date, its phase (Idea, Planning, Building, Shipped) and, for the one in progress, what is Done, In progress and Missing. A milestone reads “4 of 6 done”, and its bar fills only with finished items.
  • Goals: each goal with its KPI’s latest number.
  • Needs you, Next up, To look at (recent outputs), Lea’s latest update, Decisions, Open questions, Implemented (shipped milestones) and Later.

There is no project-wide percentage. The plan grows as you learn, so one number would mislead; the milestones answer “how far along”.

The chat with Lea. Everyone in the project sees the same chat, and every message carries its author’s name; messages sent from your own Claude Code or Codex session over MCP say “via Claude Code” or “via Codex”.

  • Lea writes what you decide into the records it changes, in the same turn. Under her reply, a Saved line lists what the reply changed, with chips that link to the records.
  • Turns where Lea only worked fold into one line (“While you were away · 3 updates”), and the steps inside a reply into “Worked · 5 steps”. Her words are never folded.
  • When she changed something on her own (at the Act, then tell me level), her message offers Undo.
  • The chip in the chat header shows her autonomy level and switches it in one tap.

The plan has three layouts, kept in the URL so you can share a link:

  • List (the default): items grouped by milestone, with chip filters. Each milestone group has a Later · Next · Now control. Moving a milestone to Now starts its work.
  • Canvas: the dependency graph. It opens on the current milestone and shows what blocks what.
  • Outputs: every output and pull request, newest first, by milestone.

People, agents and their workers, grouped by area, each with Doing now · Up next · Waiting on, one plan line per area, and the machines. A worker that has no runner to go to reads “W-1 waiting for a runner”.

One list and one count, grouped by verb:

Group What it holds
Decide Plans and brief edits
Answer Questions and stuck work
Review Screenshots, reports, designs
Allow Workers asking to do something (permission prompts)
Do Your own items
Sort Things you captured that have no place in the plan yet

The same count shows on the tab, the phone badge, the Overview and the page title. See Needs you for what each card does.

Every item has its own page: the description (its spec), its checks, its runs, outputs, thread and history. Item pages open inside the project, with ← Back to where you came from.

Settings What you change there
Account (your avatar, on every page) › Profile, Sign-in & security, Connected apps Your name, password, passkeys, sessions, and the MCP clients that act as you
Workspace › General, Members, Runners, Model providers The workspace name, invites and roles, your paired runners and their profiles, the Anthropic and OpenAI API keys
Project › General, Repositories, Areas, Lea, Agents, Workers The project’s name and status, the repos workers check out, areas, Lea and other agents (where they run, autonomy, pause), worker presets. The brief is edited on the Overview.

Claude Code is a trademark of Anthropic; Codex of OpenAI. Minga is not affiliated.

PrivacyImprintTerms