Tooling intent · separate model, client, protocol, server, and engine
Plan a Claude Opus 5, Claude Code, and MCP workflow for Unreal with honest integration boundaries, least privilege, exact tools, review gates, native validation, and SEELE handoff.
Claude Opus 5 is a model; Claude Code is a coding-agent client; MCP is a protocol; an Unreal MCP server or plugin is a separate implementation. Do not call the combination an official Unreal integration unless Anthropic or Epic publishes one. Inventory the exact server, tools, transport, permissions, project version, and failure behavior, then validate every edit or editor action in the native Unreal project.
Choose a concrete prompt, open the SEELE Unreal creator, review the prefilled brief, and generate the native project. The first prompt is tailored to this page; the remaining prompts are reusable starting points.
Create a native Unreal 5 cooperative-style puzzle that remains fully playable solo: the player routes energy between three stations, opens a vault, collects an artifact, and exits. Make state changes visually obvious, prevent invalid sequences, support reset, and structure the project so interaction, puzzle state, and level presentation have clear ownership. Provide a browser-preview path under five minutes.
Use this prompt in SEELEUse Unreal Engine to generate a simple third-person tour game with mountains and water. Include responsive movement, a clear route, one completion state, failure recovery, immediate restart, browser preview, and a native downloadable Unreal 5 project.
Use this prompt in SEELEBuild a compact dungeon crawler using Unreal Engine. Include third-person movement, one combat loop, two connected rooms, one key-and-door objective, visible health and feedback, death, completion, immediate restart, browser preview, and a native downloadable Unreal 5 project. Test and regress the result.
Use this prompt in SEELECreate a native Unreal 5 third-person city tour with a compact walkable district, vegetation, landmark lighting, and sunny, cloudy, and rainy weather states. Include clear controls, browser preview, stable restart behavior, packaging checks, and local project download.
Use this prompt in SEELEAn inspectable Unreal project rather than a model answer or an unofficial integration claim.
A fast way to check camera, controls, objective clarity, feedback, completion, failure, and restart before local continuation.
A workflow for performance review and package preparation before external distribution.
A local project handoff for source, Blueprint, asset, plugin, rights, build, target-device, and release review.
The model interprets context and decides what to propose or which approved tool to call; it does not itself provide an Unreal connection.
Claude Code or another client manages conversations, files, commands, permissions, checkpoints, and the presentation of MCP tools.
MCP standardizes discovery and invocation, but each server defines its real tools, trust boundary, transport, authentication, version support, and maintenance quality.
Unreal Editor, source, assets, build tools, automation, packaging, source control, and project-specific plugins remain the authority for the game.
Record the Opus model ID, client version, MCP server repository and revision, plugin or bridge version, transport, launch command, credentials, Unreal version, target project, and every exposed tool.
Separate read-only inspection, source writes, asset changes, editor actions, builds, tests, network calls, uploads, publishing, and destructive operations. Default to the smallest useful set and require confirmation at irreversible boundaries.
Use a clean sample project to test discovery, parameter validation, timeouts, cancellations, stale handles, editor restart, malformed outputs, concurrent calls, large assets, and recovery before touching production work.
Follow request, model decision, client permission, MCP call, server action, Unreal state change, output, native verification, and rollback. Preserve logs at every boundary so failures are attributable.
For users who want the game rather than infrastructure, use SEELE AI to generate and preview a native Unreal 5 project directly. Adopt a custom Claude Code/MCP stack only when its engineering control justifies the added operational surface.
A working npm package or MCP handshake does not prove an Unreal tool is correctly bound to the project or editor.
Tool names can conceal broad shell, filesystem, editor, network, or publishing authority.
Plugin and server versions can drift from the engine version and silently change schemas or behavior.
Screenshots and success strings are not sufficient evidence that the intended asset, source, build, or packaged state changed.
| Dimension | What good looks like | Evidence to keep |
|---|---|---|
| Discovery | Exact tools and schemas are visible and versioned | Manifest, server revision, and live capability listing |
| Safety | Each operation has least privilege, confirmation, audit, timeout, and cancellation | Adversarial permission and malformed-input tests |
| Correctness | Tool output matches the real Unreal state | Independent source, asset, editor, build, and runtime inspection |
| Recovery | Client, server, editor, and project recover from partial failure | Kill, restart, stale-session, and rollback tests |
Document the server identity, transport, executable or endpoint, credentials, tool schema, read and write targets, network destinations, confirmation rules, timeout, logging, and revocation path. A friendly tool name is not a security or correctness boundary.
Repository search, log reads, asset inspection, build status, and editor state can often remain read-only. File edits, Blueprint changes, package installation, source-control actions, builds, packaging, cloud operations, and publishing should require progressively stronger approval and isolation.
A successful tool call may still target the wrong workspace, branch, map, asset, configuration, account, or build. Cross-check returned paths and identifiers, inspect diffs and dirty assets, and run native tests before trusting the orchestration transcript.
Use the model research to tighten the brief, not to replace project evidence. The direct creation path is deliberately short and observable:
Keep one camera, one primary verb, one objective chain, explicit failure and completion, restart behavior, visual direction, controls, target session length, and a clear cut list.
Open the canonical Unreal creator, choose the closest verified starter world, submit the brief, and generate a native Unreal 5 project rather than treating a text answer as the deliverable.
Play the result from start through success, failure, and restart. Check camera, input, objective clarity, feedback, interaction state, visual hierarchy, and obvious performance or stability problems.
Review the project, source, Blueprints, assets, plugins, rights, configuration, performance, saves, networking, cook, package, and target requirements before local continuation or external release.
Create a native Unreal 5 cooperative-style puzzle that remains fully playable solo: the player routes energy between three stations, opens a vault, collects an artifact, and exits. Make state changes visually obvious, prevent invalid sequences, support reset, and structure the project so interaction, puzzle state, and level presentation have clear ownership. Provide a browser-preview path under five minutes.
Snapshot date: July 25, 2026. Anthropic's July 24, 2026 announcement is the first-party source for release status, pricing, positioning, and launch availability. It does not claim an Epic-supported Unreal integration. Epic documentation and the exact project remain authoritative for engine behavior, compilation, assets, tests, cooking, packaging, and target-platform results.
Release date, claude-opus-5 model ID, coding and agent positioning, pricing, Fast mode, alignment, safety, and availability.
Open official announcementRe-check current model behavior, API surface, prompting guidance, limits, effort settings, and migration notes before production use.
Use version-specific Epic documentation and the real project for C++, Blueprint, build, test, cook, and package decisions.
Calculate Claude Opus 5 cost for Unreal coding and agent tasks using official $5/$25 token pricing, effort, cache behavior, tools, retries, review time, and task-level budgets.
Read this guideCompare Claude Opus 5 and GPT-5.6 for Unreal coding, debugging, agents, computer use, cost, safety, native validation, and task routing without unsupported winner claims.
Read this guideThreat-model Claude Opus 5 Unreal agent workflows across source, assets, tools, prompts, credentials, builds, plugins, publishing, data handling, approvals, and SEELE handoff.
Read this guideNo such inclusion is established by the cited release. The model, client, protocol, MCP server, and Unreal integration are separate layers.
Claude Code is a coding-agent product, not an Unreal Editor plugin. A third-party bridge may connect them and must be evaluated separately.
Begin with read-only project inspection, then add narrowly scoped writes or editor actions only after tests, audit, confirmation, and rollback are proven.
Inspect the actual file, asset, editor state, build output, runtime behavior, and source-control diff rather than trusting the tool response alone.
When the goal is to generate a native Unreal project and playable result rather than maintain a custom agent-to-editor integration stack.
Log the model and client, server and tool identity, workspace and revision, parameters, permissions, confirmations, outputs, changed files or assets, native test results, failures, and rollback.
Open the canonical SEELE Unreal creator, choose a verified starter world, generate the native project, inspect the browser preview, then download or continue optimization and packaging. Keep third-party model evaluation and project evidence separate.