# Ue Agent Bridge

> Give coding agents eyes and hands in Unreal Engine with a plugin, MCP bridge, and skills built to cut token-heavy guesswork

- **Type:** MCP server
- **Install:** `agentstack add mcp-aborigen19961202-ue-agent-bridge`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [aborigen19961202](https://agentstack.voostack.com/s/aborigen19961202)
- **Installs:** 0
- **Category:** [Integrations](https://agentstack.voostack.com/c/integrations)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [aborigen19961202](https://github.com/aborigen19961202)
- **Source:** https://github.com/aborigen19961202/ue-agent-bridge

## Install

```sh
agentstack add mcp-aborigen19961202-ue-agent-bridge
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# UE_AgentBridge

`UE_AgentBridge` gives coding agents eyes and hands in Unreal Engine.

It combines an Unreal plugin, an MCP bridge, and companion agent skills so tools like Codex and Claude can look inside a live Unreal Editor session, understand what is actually happening, and perform a small set of safe actions.

The point is to stop wasting time and tokens on guesswork. Instead of making the model infer Unreal state from code alone, `UE_AgentBridge` lets it ask the editor directly while code edits, builds, and git stay in the normal repo workflow.

This project is plugin-first:

- TypeScript MCP server over stdio
- reusable Unreal project plugin under [`unreal-plugin/UEAgentBridge`](./unreal-plugin/UEAgentBridge)
- `plugin` mode for normal live-editor use
- `remote-control` fallback for asset search and explicit property reads/writes
- `mock` mode for local validation without Unreal

## Why It Exists

AI agents can write and change Unreal code, but they usually cannot see what Unreal Editor is doing right now.

This project fixes that gap.

With `UE_AgentBridge`, an external agent can:

- check whether the editor is ready and connected
- inspect selected actors, level actors, and object properties
- make narrow, explicit property changes
- read output log slices and editor diagnostics
- move the viewport camera and capture useful screenshots
- inspect debug draw output for visual debugging
- check Live Coding readiness and trigger a safe reload path
- run only a small allowlisted set of safe console actions

That helps in two ways:

- the agent can query real editor state instead of guessing from code alone
- it can pull the exact screenshot, actor state, or log slice it needs instead of burning tokens on back-and-forth trial and error

What stays outside Unreal on purpose:

- repo navigation
- file editing
- git
- shell and build command execution
- architecture and code reasoning

Unreal provides the eyes and hands. The repository stays the source of truth for code work. That split is what makes the workflow safer, clearer, and cheaper in tokens.

## Install

Requirements:

- Node.js 20+
- Unreal Engine 5.7 tested on Windows

Install JS dependencies:

```bash
npm install
```

Install the Unreal plugin into a target project:

1. run `powershell -ExecutionPolicy Bypass -File .\scripts\sync-ue-plugin.ps1 -ProjectRoot ""`
2. build `Editor`
3. launch Unreal Editor

Build a distributable plugin zip:

```bash
npm run package:plugin
```

That stages a clean plugin package under `artifacts/plugin-package` and creates a zip for third-party installation testing.

## Quick Start

### Local mock validation

1. `npm install`
2. `npm test`
3. `npm run dev`
4. run `ue_healthcheck`

### Real Unreal validation

1. install the `UEAgentBridge` plugin into the target Unreal project
2. make sure Unreal Editor is running locally
3. keep Remote Control enabled for `ue_asset_search`, `ue_get_property`, and `ue_set_property`
4. start the bridge with `UE_BACKEND_MODE=plugin`
5. run `ue_healthcheck`
6. exercise a few live tools such as `ue_get_level_actors`, `ue_get_output_log`, `ue_get_editor_state`, `ue_frame_actor`, and `ue_get_live_coding_status`

### Build, run, and checks

```bash
npm run typecheck
npm test
npm run build
npm run dev
```

For repeatable plugin smoke validation:

```bash
npm run build
npm run smoke:plugin
```

For C++ iteration, use the external build loop in [CPP_ITERATION_WORKFLOW.md](./docs/guides/CPP_ITERATION_WORKFLOW.md). Full external `Editor` target builds should run with Unreal Editor closed, while in-editor iteration should use the bridge's Live Coding path after a successful external build.

## Architecture

The repository is split into:

- MCP layer: tool registration, validation, formatting, and error handling
- backend layer: Unreal transport adapters
- Unreal plugin package: bounded editor-side backend
- docs: operator guides, release notes, contracts, and archive material

Repo layout:

```text
docs/
  archive/
  contracts/
  guides/
  integrations/
  planning/
  release/
src/
  backend/
  config/
  server/
  tools/
  types/
  utils/
test/
unreal-plugin/
  UEAgentBridge/
```

## Contributing

Contributions are welcome.

Start with [CONTRIBUTING.md](./CONTRIBUTING.md) before opening a pull request.
For workflow-sensitive changes, keep the repo/Unreal boundary aligned with [AGENT_PLAYBOOK.md](./docs/guides/AGENT_PLAYBOOK.md).

## License

This project is released under the [MIT License](./LICENSE).

## Documentation

Most users only need the core guides:

- [AGENT_PLAYBOOK.md](./docs/guides/AGENT_PLAYBOOK.md)
- [CPP_ITERATION_WORKFLOW.md](./docs/guides/CPP_ITERATION_WORKFLOW.md)
- [PLUGIN_RUNBOOK.md](./docs/guides/PLUGIN_RUNBOOK.md)
- [THIRD_PARTY_INSTALL.md](./docs/guides/THIRD_PARTY_INSTALL.md)
- [CLIENT_INTEGRATION.md](./docs/guides/CLIENT_INTEGRATION.md)

Full documentation index:

- [docs/README.md](./docs/README.md)

Useful starting points:

- [AGENT_PLAYBOOK.md](./docs/guides/AGENT_PLAYBOOK.md) for the intended repo/Unreal operating model
- [PLUGIN_RUNBOOK.md](./docs/guides/PLUGIN_RUNBOOK.md) for day-to-day plugin workflow
- [THIRD_PARTY_INSTALL.md](./docs/guides/THIRD_PARTY_INSTALL.md) for installation outside this repo
- [FAB_RELEASE_READINESS.md](./docs/release/FAB_RELEASE_READINESS.md) for marketplace preparation

Internal planning, contracts, release checklists, and historical notes remain available from the full docs index.

## Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [aborigen19961202](https://github.com/aborigen19961202)
- **Source:** [aborigen19961202/ue-agent-bridge](https://github.com/aborigen19961202/ue-agent-bridge)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/mcp-aborigen19961202-ue-agent-bridge
- Seller: https://agentstack.voostack.com/s/aborigen19961202
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.