<p align="center"><img src="docs/images/logo.jpg" alt="Dream Bot Team for UEFN: AI agents for Fortnite maps" width="640"></p>

UEFN Dream Bot Team: AI agents and MCP tools for Fortnite maps

Dream Bot Team is a free, open-source Claude Code kit for building, testing and improving UEFN (Unreal Editor for Fortnite) maps with AI. It gives you twenty-two specialized AI agents (seventeen team agents plus five UEFN MCP sub-agents), each with one narrow job, instead of one general-purpose assistant trying to do everything and stepping on its own toes. It writes and fixes Verse code in a plan-first loop, and it drives the UEFN MCP (Epic's Model Context Protocol server for the editor) with recipes that were proven on a live editor.

• AI agents with a plan-first workflow: no task counts as done until two independent reviewers sign off.
• Skills for Verse, level devices, UI widgets, Niagara VFX, materials, physics, performance and map growth (titles, thumbnails, trailers, launch plans).
• Every UEFN MCP tool mapped, recipes proven on a live editor: arguments, prerequisites, cleanup and known traps for each tool.
• Learns from your maps and keeps your data on your computer; sharing lessons is optional and private by default.
• Agent Console: a local dashboard that shows each agent working, MCP activity and token use.
• MIT license, works for one island or fifty.

Built by someone juggling ~50 UEFN islands at once.

Read the manual: quick start, the team, the console, the UEFN MCP tool reference, troubleshooting.

Screenshots

The Agent Console header — session stats, token usage, and the "miniflow" rail tracking a task live through its five gated stages (intent-gate → coder → intent-reviewer → compliance-reviewer → planner-docs):

The full Agent Console — the header above, plus every agent's live status at a glance (demo data):

The Skills page — what the kit has learned across all your maps (top), and what it learned from the map you are working on (below): learning steps, the map blueprint, a log of what it read, the patterns of the genre and the lessons waiting for your approval. Sample data:

The Docs page — roadmap progress, bugs by severity, the task timeline and the owner actions, read from the project's own documents (demo data):

The Flow page — the pipeline as a U-shaped board: one card and one agent icon per station, tokens per agent, Control Tower and Pipeline pulse (demo data):

The full pipeline — every agent and every hand-off, end to end:

What's new (v1.102 - v1.108.4)

• The kit updates itself (1.108.3). A background check of the newest GitHub release runs at session start, at most every 6 hours, and downloads a newer one into your profile; the project follows at its next session. If you have an older version, do this once: run /kit-update --fetch in a session (or update-my-profile.cmd from the repo), then open a new session. Older kits cannot update themselves because nothing in them looks at GitHub. Off switch: an empty file ~/.claude/kit-auto-update.off.
• Score page (1.108.1). A new console tab that measures the project against its own plan, release by release: a chart of plan, built and proven with a forecast, and a green, yellow or red light per criterion so you see where to act first. It reads ROADMAP, BUGS and the readiness verdict by itself; new criteria are proposed on the Skills page and need your approval.
• Ready for the jam (1.108). A new project asks whether it is a map or a Fab product (a map gets the kit, a Fab product never does). Verse headers keep only the present and every change goes to a log. /concept-card scores an idea, /devlog writes the dev log from the logs, /jam-check checks the entry before you hand it in.
• The MCP, mapped, learned and proven (1.107). The Console is now a small solar system: five planets around the chief of staff, the other agents as moons, token arcs, a detail card, pinned agents and switches to show or hide each part. 14 UEFN MCP skills and 5 MCP sub-agents ship in the kit; recipes are marked tool-ok or proven, and what you prove can be shared with the community (counts only). UEFN can be restarted after a crash if you opt in; a running editor is never closed.
• Flow page as a U-shaped board (1.106). Twelve stations along a U, one card and one agent icon per station, zone progress bars, the Control Tower and a Pipeline pulse. Tokens, time and verdicts come from the real log and /agent-costs; a missing number is shown as a dash, never invented.
• Tokens per agent (1.104). A table at the bottom of the Console, counted from the sub-agent transcripts: runs, calls, input, output and cache, per agent. Counts, not prices.
• The kit maps UEFN's MCP by itself (1.106). All 388 tools in 30 toolsets are catalogued at session start (no AI tokens), safe tools are test-called, and the Skills page shows how much of the MCP the kit really uses. New chapter in the manual: "The MCP engine". An optional laboratory mode (mcp_lab.py, with its own safety checks) lets agents try every tool in a disposable project.
• Verse UI Kit skill (1.106). verse-ui-kit teaches in-game screens with or without Epic's free kit; /ui-kit-check tells whether it is installed and whether Epic changed it. Epic's guide is never copied into this repo.
• Verse build check (1.102 - 1.103). /verse-check asks the open UEFN to compile and reports clean, errors or not analysed; build errors are logged and recurring ones become lesson proposals. /text-scan finds emoji and invisible characters in publication text.
• Fewer questions, easier handovers (1.105). Pre-approved read-only commands and /handoff for long sessions.
• Files are parked, not lost (1.106). If a delete is refused, the file goes to _to_delete and you are told; the Docs page now follows the files by itself.

Full list: CHANGELOG.md.

What's new (v1.100 - v1.101)

• A manual you can read in the browser: 13 chapters (quick start, full installation, the team, the console page by page, skills and lessons, community, troubleshooting, FAQ, glossary) with screenshots. Online at mimmo-the-root.github.io/UEFN-dream-bot-team, and built into the kit at http://127.0.0.1:8765/manual/.
• A "Help for this page" link and a footer on every console page (repo, MIT license, author), and the project logo.

What's new (v1.99)

• Share your lessons with the community (voluntary). A switch on the Skills page and a Contribute button on each approved lesson: you see the exact JSON, then GitHub opens a pull request from your own account. Only lessons, only from a published map; map names, code and blueprints never leave. A curator agent proposes what to promote and the maintainer validates it before a release; creators receive the community lessons as proposals to approve. See docs/community-contributions.md.
• Safe by design. Fixed format, screening against instructions aimed at an assistant, an isolated classifier, a strict plain-English character filter, statements regenerated by the curator, revocation of a lesson from everyone's skills.
• New page: What's new. The last release, new and improved skills, top contributors, thanks and a "Submit an issue" link.

What's new (v1.98)

• The kit learns from your maps by itself. At the first session on a map it reads the Verse code, compares your project documents with the code (and corrects the ones that are wrong), reads the devices placed in the level, their wiring and settings, and the island configuration through the UEFN MCP. It never interviews you, and if the session stops it resumes where it stopped.
• Map blueprint. What it learned becomes a blueprint inside the genre skill: how a map of this kind is built (layout with counts, game loop, devices and how they are wired, rules with the numbers observed, build order). It follows your development: when code, documents or devices change, the blueprint is marked out of date and rewritten at the next session. Items it could not read are retried at the next starts (never invented), and every analysed map also queues 3 to 5 general lessons for your approval.
• You confirm genre and map type before anything is saved. The genre is one of Epic's official Discover genres; the map type is a free name (pillars, red-vs-blue, checkpoint-race...). The kit only proposes them: you confirm or correct both on the Skills page, and until then nothing touches your skills. Confirming saves the waiting blueprint and lessons by itself.
• IP variants. A map that is the same base game dressed for an IP is a type of the same genre: the kit reads the IP from your documents, proposes it as the type (the shared base is base), and the blueprint of the variant records only what differs from the base.
• Skills page rebuilt: the kit overview on top (waiting lessons, maps, genres and types, confidence, privacy), the genre of this map below, Approve all for waiting lessons, and the SKILLS tab blinks on every console page when something waits for you.
• **Do