Title: The platform: BUD board, forecasts, XP and People | Glideship
Source: https://glideship.ai/platform/
Summary: The BUD board, AI-PERT delivery forecasts, quality loops, knowledge synced from your code, skill profiles, XP, and leave and HR records in one self-hosted app.

See it

# The platform

The BUD board, delivery forecasts, quality loops, living knowledge, XP and the People workspace. Everything your team works in, in one place.

The BUD board. Every feature, from the first request to production.

Watch it run

## From install to code graph.

Set Glideship up on your own machine, then watch it scan your repositories into a map of what your code does.

[Watch with sound 2:01](https://www.youtube.com/watch?v=Y1y_CwpmnQs)

Repository scans and the code graph — How Glideship reads your repositories and maps what your code does.

**Read the transcript**

This video shows how Glideship reads your code. Repositories are connected here, under code and repositories in settings. Each repository has its main, develop and UAT branches mapped. More can be added from a local folder, a GitHub clone, or a bulk import. A scan runs in stages for each repository.

It indexes the code into a call graph, groups related code together, and works out who knows which parts. It also pulls out the design system and the backend routes, and writes up the features it finds. Each stage can be opened to see what it did. On a later scan, a repository that has not changed is skipped.

At the end, one more step matches the API calls made by the frontend with the routes in the backend, so a feature is linked across repositories. The feature registry lists what the product already does, mapped from the code. Each feature shows where it lives, which backend it calls, the BUD it was planned in, and who touched it last.

The graph view draws every repository and its features in one map, with the links between them. Turn on bus factor, and features that only one or two people have worked on are ringed. Skill profiles are read from git history and refreshed with each scan. By area, you can see who knows each part of the code, and which parts only one person knows.

When a pull request merges, Glideship re-indexes that repository and refreshes only the features the change touched. The map stays current without anyone maintaining it.

[Watch with sound 1:31](https://www.youtube.com/watch?v=I9Sfb9lmcPQ)

Setting up Glideship — From an empty install to the first scan of your code.

**Read the transcript**

This is a fresh Glideship install. The setup wizard takes you from an empty database to a first scan of your code in six short steps. First, name the workspace. Ours is Northwind Labs, and the URL slug fills itself in. Next, the first admin. Kavya will own this workspace.

Then the AI engine. Glideship runs its agents through Claude Code, GitHub Copilot, OpenAI Codex, or a local Ollama model. Each organisation picks one. Here we reuse the Claude login that is already on this machine. You could paste an API key instead. The test checks that the agent can actually run.

Now the code. You can clone from GitHub, import in bulk, or point at a checkout on disk. We add the two TaskFlow repositories and map their main and develop branches. When we continue, the repositories are registered and the first scan starts in the background.

The dashboard follows the scan as it runs, with a short checklist of what is left to set up. Glideship is open source and self-hosted, so your BUDs and scan results stay in your own database.

One place per feature

## The BUD holds the whole feature.

Every feature lives in one BUD, a Business Understanding Document. It holds the spec, tech spec, test plan, acceptance criteria and full history, and replaces scattered tickets, docs and wiki pages.

BUDDesignTech archDevelopmentTestingUATProdClosed

- Full history: stage changes, owners, reopens and bugs

- Searchable by meaning, so every agent can find it

Delivery forecasts

## Dates with odds, not points.

Story points and planning poker are gone. AI gives a best-case, likely and worst-case estimate for each phase (PERT). Then 10,000 Monte Carlo simulations turn those into dates that are 50%, 70% and 85% likely.

Each developer's skills, the backlog and current workload are taken into account. Forecasts get sharper as you ship, because the Learning Agent compares every estimate with what really happened.

Example forecast

Notification redesign

Complexity 3 of 5

Alice

Backend0.92Frontend0.35

50% likely byApr 22

70% likely byApr 25

85% likely byMay 2

An illustrative forecast: the spread of 10,000 simulated finish dates, with the dates that are 50%, 70% and 85% likely.

Quality and backlog

## Bugs go back where they came from.

Bugs are matched to the feature they came from and routed before they pile up. The backlog is ordered by data, not gut feeling.

### Quality loops

- Bug Limit
  Complexity × a multiplier you set for your organization. Go over it and the work is reassigned automatically.

- Auto-Reassignment
  The original developer moves to bug review, and QA moves on to the next waiting BUD.

- Feature Reopening
  Bugs from the outside world reopen the Feature and send it back through triage.

- Auto-Labelling
  Each bug is labelled "missed feature" or "development bug", and each takes a different fix path.

- Knowledge Capture
  Every bug fix goes into the knowledge base, so the same kind of bug is less likely to come back.

### Backlog order

- Capacity-Aware Triage
  The Triage Agent lowers the priority of new work, or holds it back, based on how much room the team has right now.

- Dynamic Reassignment
  The Reassignment Agent moves work around as business needs shift.

- Customer Priority Scoring
  Customer revenue (ARR), severity and customer tier set the backlog order automatically.

- Live Workload View
  Tracks how busy each developer is, so work stays balanced.

Living knowledge

## Knowledge that keeps up with the code.

Four layers of knowledge replace the stale wiki. They stay in sync on their own, not on a timer, and every agent reads from them.

- Synced from the code, not kept up by hand

- Search by meaning, not just keywords

- Re-learned every time a pull request merges to your main branch

- Fed straight to the agents, so they always work from the latest

1. 1
  Your code
  Source code plus a CLAUDE.md guide for each repo. Re-learned on every merge to your main branch.
  Every merge to main

2. 2
  Team standards
  Coding standards, design guidelines and API patterns. Syncs when they change.
  When they change

3. 3
  Decisions
  BUDs, your team's rules and architecture decisions. Updated in real time.
  Real time

4. 4
  Search by meaning
  Find anything by what it means, not just the exact words. Indexed automatically.
  Automatic

Industry research

61%

of professional developers spend more than 30 minutes a day searching for answers or solutions to problems. — Source[Stack Overflow Developer Survey 2024](https://survey.stackoverflow.co/2024/professional-developers)· 28,911 responses

Skill profiles

## Profiles that keep up.

The Skill Agent rebuilds developer profiles every day. No forms to fill in, no out-of-date CVs.

- Daily Profile Rebuilds
  Reads git history, BUD work and bug fixes to score each skill from 0 to 1.0, module by module.

- Bus Factor Alerts
  Flags modules that only one person has worked on, before that knowledge walks out the door.

- Assignment Suggestions
  Suggests developers for new BUDs by matching skills with free time.

The XP layer

## Levels, badges and a leaderboard.

Skills grow, badges unlock, and the leaderboard shows what people shipped, not how many tickets they touched.

The People workspace

## Leave and records, in the same place.

The rest of the company works here too. Leave past someone's balance becomes loss of pay instead of a refusal, any of their managers can approve it in Slack, and each location keeps its own holidays. Bank and ID numbers are encrypted with a separate HR key, and every reveal is logged.

[See the People page](https://glideship.ai/people/)

## Explore it for yourself

[Get started on GitHub](https://github.com/mickyarun/glideship)

One command. Needs Docker and Node.js 18 or later. Your code and data stay on your machine. `npx glideship init`
