The Delegation
Design, run, and inspect Gemini multi-agent teams in a no-code 3D office.
- Source repo
- arturitu/the-delegation
- Stars
- ★ 660
- Last updated
- 5mo ago
- License
- NOASSERTION
- Primary language
- TypeScript
- FA score
- 34/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specific
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- An educator teaching multi-agent collaboration can use a predefined team and the 3D office to make roles, handoffs, and task flow visible in a lesson.
- Not a fit if
- Teams avoiding vendor lock-in; models and multimodal generation are Gemini-only
- Commercial projects needing the 3D assets, which are CC BY-NC 4.0 non-commercial
- Source review
- 34/100 · Major gaps 8 safety controls not found
What does this agent do, and when should you use it?
The Delegation is a no-code 3D environment for AI enthusiasts, educators, and creative developers exploring multi-agent collaboration. It combines a shared office rendered with Three.js WebGPU and TSL with a React Flow node editor for team structures and task flows. Users can start from six predefined industry-team templates, create teams, assign Gemini models to individual roles, and inspect flows through visualizers, logs, and an agent inspector. Agent teams can generate text, images, music, and video through the Gemini API, while the interface estimates token use and costs. A hosted full experience is linked, and the repository documents local development via npm; operating the simulation requires a user-supplied Gemini API key.
A user creates multi-agent designs in the React Flow-based Team Editor or begins with predefined teams such as Creative Agency, Film Studio, and PR Agency. Individual roles can be assigned different Gemini Models, and teams generate text, image output with Nano Banana, music with Lyria 3, and video with Veo 3.1. The Team Flow Visualizer shows hierarchy and task flows, while Kanban & Action Logs and Technical Logs expose progress, raw LLM traces, tool calls, and structured agent responses. The embodied layer uses Three.js WebGPU, a NavMesh, and three-pathfinding so NPCs navigate to and claim Points of Interest including desks, seats, and computers, transitioning among walking, sitting, working, and talking. In human-in-the-loop workflows, Simulated PR Reviews let the user review proposals, provide feedback, and merge tasks.
- An educator teaching multi-agent collaboration can use a predefined team and the 3D office to make roles, handoffs, and task flow visible in a lesson.
- A creative developer designing an AI workflow can rearrange roles in Team Editor and assign Gemini models according to speed or reasoning needs.
- A content team experimenting with role-based generation can have an agent team produce text, Nano Banana images, Lyria 3 music, or Veo 3.1 video.
- An AI enthusiast who needs execution visibility can inspect an agent's mission, thoughts, history, action logs, and technical traces.
- A team practicing human approval in an agent workflow can use PR-style workflows and Simulated PR Reviews to review, comment on, and merge proposals.
How do you install or deploy this agent?
Obtain a Gemini API key first: the simulation requires BYOK. Clone or fork the repository, then run these commands in the repository directory:
npm install
npm run devOpen the local URL printed by the terminal; the documented example is usually http://localhost:3000/the-delegation. The documentation does not specify the local API-key configuration mechanism, environment-variable name, Node.js version, or a production deployment procedure, so the key must be supplied through the interface or project configuration supported by the implementation.
How do you use this agent?
Start the development server, open the local app, and provide a Gemini API key to run the simulation. Choose a predefined team or create a node-based structure in Team Editor, assign Gemini Models to roles, and run the team flow. Use Team Flow Visualizer, Agent Inspector, Kanban & Action Logs, and Technical Logs to inspect activity. When a human-in-the-loop approval occurs, use Simulated PR Reviews to review the proposal, give feedback, or merge the task.
What are this agent's strengths and limitations?
- It combines a React Flow team editor, live flow visualization, and a Three.js WebGPU office in one environment, making multi-agent coordination tangible.
- Per-agent Gemini model assignment is paired with token and cost estimates plus visibility into raw LLM traces, tool calls, and structured responses.
- It supports four output types—text, Nano Banana images, Lyria 3 music, and Veo 3.1 video—rather than limiting teams to text.
- NavMesh-based NPC movement with three-pathfinding gives task state a spatial representation through desks, seats, and computers.
- Its core AI provider is the Gemini API and the full experience requires a user-provided Gemini API key, so it is not provider-agnostic.
- Local documentation does not provide API-key setup details, environment-variable names, a Node.js version, or a production deployment process.
- The roadmap still marks the space editor, long-term team memory, deeper embodied AI, and further architecture decoupling as unfinished.
- Repository metadata is NOASSERTION while the documentation claims MIT code and CC BY-NC 4.0 3D assets; commercial adopters need to verify the applicable license scope.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| The Delegation This agent | 34 · Major gaps | Web appFree + model costs | ★ 660 | 5mo ago | TypeScript | — |
| Paper2Poster | 28 · Major gaps | CLIFree + model costs | ★ 3.9k | 3mo ago | Python | Codex · OpenAI API |
| WorldX AI World Builder | 56 · Major gaps | Web appFree + model costs | ★ 1.5k | 23d ago | TypeScript | OpenAI API |
| Stitch Design Skills | 0 · Major gaps | Agent plugin / skillFree + model costs | ★ 8.4k | 1mo ago | TypeScript | ChatGPT · Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows a front-end demo with no backend or permission management code, so least privilege, user confirmation, data flow transparency, sensitive data handling, external effects, and rollback are all absent, scoring 0. Dependency security: dependencies in package.json are not fully pinned (e.g., ^), and no security audit or vulnerability scanning evidence is provided, scoring 0. Source attribution: author info is clear in README and package.json, but publisher is unverified and no contributor or maintainer info is present, scoring 0.
Self-consistency: README features align with package.json dependencies and scripts, no contradictions, scoring 2. Dependency availability: dependencies are well-known libraries, but versions are not fully pinned, posing potential update risks, scoring 2. Failure messages: no error handling or user notification info provided, scoring 0.
Audience and scenarios: README clearly targets AI enthusiasts, educators, and developers, and provides multiple usage scenarios, scoring 2. Capability boundaries: README lists features and roadmap but does not explicitly state unsupported features, scoring 2. Trigger precision: no trigger mechanisms or command descriptions provided, scoring 0. Environment fit: installation and run instructions provided, supports local and online deployment, scoring 2.
Information architecture: README is well-structured with intro, features, tech stack, roadmap, etc., scoring 2. Install notes: provides npm install and npm run dev, scoring 2. Naming stability: project name and version are clear, scoring 2. Examples and FAQ: README provides feature examples but no FAQ, scoring 1. Known limitations: no explicit known limitations listed, scoring 1. License: dual license MIT and CC BY-NC 4.0 provided, scoring 2. Versioning and changelog: version number present but no changelog, scoring 1. Maintenance responsibility: author info clear but no maintenance commitment, scoring 2.
Output usability: project generates text, image, music, video assets, but requires user-provided API key, scoring 2. Marginal value: provides unique 3D multi-agent visualization, but similar tools may exist, scoring 2. Cost-benefit: requires user-provided API key, may incur costs, but project itself is free, scoring 1.
Claim traceability: feature claims in README lack code references or test evidence, scoring 1. Cross-source corroboration: no external sources to verify, scoring 0. Fact-inference separation: README distinguishes features and roadmap, but does not clearly separate facts and inferences, scoring 1.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- Project is a front-end demo with no backend or permission management; users must provide their own API keys, so be cautious with key security.
- Dependency versions are not fully pinned, posing supply chain risks; consider using lock files.
- Publisher is unverified; assess risks before use.