Other formats agents might prefer:
markdownjsonllms.txt

Agent? You probably want markdown or json, or Pod over MCP.

Reported issues for gopeak

Pod holds 17 of 18 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user.

Back to gopeak.

Most discussed

[Bug] Cannot run Gopeak on windows

GoPeak version: 2.3.2 Godot version: 4.6.1 Node.js version: 22.x OS: Windows

MCP Client: OpenCode

Description

The npx command returns an empty output, and trying to install globally results in errors.

Steps to Reproduce

  1. Open PowerShell
  2. Enter "npm i -g gopeak"
  3. Watch the errors in despair

Expected Behavior

For gopeak to install globally

Actual Behavior

Errors: ` npm warn cleanup Failed to remove some directories [ npm warn cleanup [ npm warn clea

Read the thread · 2026-03-05 · closed · external user · 3 comments

[Feature] 95 mcp tools is a token nightmare

Problem

What problem does this feature solve?

95 Mcp tools will waste tons of token.

Proposed Solution

Describe the feature or tool you'd like to see.

Apply dynamic tool discovery

Alternatives Considered

Any alternative solutions you've considered.

Additional Context

Any other context, screenshots, or examples.

Read the thread · 2026-02-22 · closed · external user · 3 comments

[Bug] What am I doing wrong?

GoPeak version: latest Godot version: 4.7 Node.js version: v26.2.0 OS: win

MCP Client: Antigravity

Description

1.- Dowload the repo zip 2.- Run: Npm install 4.- Put addons/ in my project godot *test/.godot *test/addons/here 5.- ACTIVE the Plugins 6.- Create mcp_config with:

{
  "mcpServers": {
    "godot": {
      "command": "npx",
      "args": ["-y", "gopeak"],
      "env": {
        "GODOT_PATH": "C:/Program Files/Godot/Godot_v4.4.1-stable_win64.exe",
       

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/67) · 2026-06-21 · closed · external user · 2 comments

### [Bug] capture_screenshot fails with validation error due to malformed ImageContent response

## Environment

- **GoPeak version**: 2.3.3
- **Godot version**: 4.6
- **Node.js version**: 22.14.0
- **OS**: macOS (Darwin 24.3.0)
- **MCP Client**: Claude Code (Claude CLI)

## Description

The `capture_screenshot` runtime tool fails with a validation error. The `resolveRuntimePayload` function in `src/index.ts` constructs the image content block with `{ type: 'image', text: parsed.data }` instead of the MCP SDK's required `ImageContent` shape `{ type: 'image', data: parsed.data, mimeType: 'im

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/19) · 2026-03-10 · closed · external user · 2 comments

### [Bug] - Two related issues: Port 7777 conflict + testing tools missing compactAlias · Issue Update

Please notice: I'm not a programmer so I used AI to help me report this bug, I hope it helps🙏

## Environment
- GoPeak version: 2.3.9
- Godot version: 4.7.stable.official.5b4e0cb0f
- Node.js version: v25.8.2
- OS: Windows 11
- MCP Client: Zoo Code

---

## Issue 1: Port 7777 Conflict - Core tools fail when runtime is already running

### Description
When the Godot game instance is already running with MCP Runtime autoload on port 7777, core tools like `project_setting_get` fail because they try 

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/74) · 2026-07-21 · open · external user · 1 comment

### [Bug] AI로 기능 테스트 좀 시켰는데 이렇게 리포트함

# gopeak Bug Reports

Repository: https://github.com/HaD0Yun/Gopeak-godot-mcp

각 이슈의 Title과 Body를 그대로 GitHub Issue에 붙여넣으세요.

---

## Issue #1

**Title:** [Bug] Vector2 property conversion fails in add_node / set_node_properties

**Body:**

### Environment
- OS: Windows 11 Pro 10.0.26200
- Godot: v4.6.2.stable.official
- gopeak: v2.3.5 (npm)
- Node.js: v24.14.1

### Bug Description
`add_node` 또는 `set_node_properties`에서 Vector2 속성(`position`, `scale` 등)을 설정하면 값이 잘못 변환됩니다.
- `scale`이 어떤 값을 넣어도 `Vec

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/47) · 2026-04-02 · closed · external user · 1 comment

### [Bug]  설치했더니 동작을 안하는데...

- OS: Windows 11 Pro 10.0.26200
  - Godot: v4.6.2.stable.official
  - gopeak: v2.3.5 (npm)
  - Node.js: v24.14.1
  - 에디터 플러그인: godot_mcp_editor v1.0.0 (build/addon에서 복사, 파일 동일 확인됨)

  증상

  - editor-status가 항상 connected: false 반환
  - scene-nodes 등 에디터 브릿지 필요 도구 사용 불가 ("Godot Editor not connected" 에러)
  - project-info 등 파일 기반 도구는 정상 작동

  확인된 사항

  1. 에디터 플러그인 정상 활성화 — Project Settings → Plugins에서 3개 모두 Enabled
  2. 에디터 UI에서 "MCP: Connected" 초록색 표시 — 에디터 플러그인은 연결 성공으로 인식 중
  3. TCP 연결 확인됨:
  TCP 

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/46) · 2026-04-02 · closed · external user · 1 comment

### [Bug] windows  install failure

**GoPeak version**: latest
**Godot version**: (e.g., 4.6)
**Node.js version**: Node.js v24.12.0
**OS**:Windows

**MCP Client**: (e.g., Claude Desktop, Cursor, Cline, OpenCode)

## Description

npm install -g gopeak

## Steps to Reproduce

1. ...
2. ...
3. ...

## Expected Behavior

What you expected to happen.

## Actual Behavior

What actually happened. Include error messages if any.

## Logs

<details>

$ npm install -g gopeak
npm warn cleanup Failed to remove some directories [
npm warn cle

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/43) · 2026-03-31 · closed · external user · 1 comment

## Most recent

### Follow up on entrypoint decomposition, opt-in setup hooks, and metadata parity

## Summary
Follow up on recent platform hardening by making shell-hook setup opt-in at install time, splitting reusable tool metadata out of the server entrypoint, and resynchronizing release-facing docs/metadata with the current GoPeak surface.

## Why this matters
GoPeak had three adjacent maintenance gaps after the recent 2.3.x platform work:

1. `src/index.ts` was still carrying a large amount of reusable static metadata, which made the entrypoint harder to reason about and review.
2. `npm i

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/40) · 2026-03-23 · closed · 0 comments

### Security: TCPServer on port 7777 bound to 0.0.0.0 with no authentication

## Summary

The runtime addon `addons/godot_mcp_runtime/mcp_runtime_autoload.gd` starts a TCPServer on port 7777 bound to **all network interfaces** (`0.0.0.0`) with no authentication. Any machine that can reach the port can connect and issue commands.

## What's exposed

An unauthenticated TCP connection on port 7777 can:

- `call_method` — call any method on any node in the scene tree
- `set_property` — modify any node property
- `get_tree` / `get_node` — read the full scene tree and node prop

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/38) · 2026-03-21 · closed · external user · 1 comment

### [Bug] inspect_runtime_tree, call_runtime_method, set_runtime_property return stubs instead of runtime data

**GoPeak version**: 2.3.3
**Godot version**: 4.6.1
**Node.js version**: 24.0.1
**OS**: macOS (Apple Silicon)

**MCP Client**: Claude Code

## Description

Runtime tools (`inspect_runtime_tree`, `call_runtime_method`, `set_runtime_property`, `get_runtime_metrics`) return stubs or process logs instead of actual runtime data from the addon, despite the addon being connected and responding to commands.

## Steps to Reproduce

1. Install the `godot_mcp_runtime` addon in a Godot project and enable it

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/35) · 2026-03-18 · closed · external user · 0 comments

### Discord link is expired

Can't discuss if can't join, please set link expiration to Never, thank you

[Read the thread](https://github.com/HaD0Yun/Doyunha-Gopeak/issues/34) · 2026-03-18 · closed · external user · 0 comments

### OpenAI-compatible tool name validation fails on invalid custom tool names

## Summary
When using Cline with an OpenAI-compatible API, requests can fail with:

```json
{"message":"400 tools.13.custom.name: String should match pattern '^[a-zA-Z0-9-]{1,128}$'","status":400}

This indicates at least one tool exposed to the OpenAI-compatible endpoint has an invalid custom.name.

Expected

Tool names sent to the OpenAI-compatible API should always be sanitized to match:

^[a-zA-Z0-9-]{1,128}$

Actual

At least one bridged tool name contains unsupported char

Read the thread · 2026-03-12 · closed · 0 comments

Follow up on the CLI prototype for benchmarkable command surfaces

Issue: Follow up on the CLI prototype so benchmarkable command surfaces are explicit, reproducible, and maintainable

Summary

The repository now contains a narrow CLI prototype and benchmark helpers for the GoPeak CLI-vs-MCP evaluation, but the current surface still mixes validated benchmarkable paths with prototype-only or partially duplicated paths. We should harden the CLI prototype around the command families that are already useful for fair comparison, make unsupported/non-comparable f

Read the thread · 2026-03-12 · closed · 1 comment

Make scene.create benchmarkable via a plugin-backed editor lane or shared execution path

Issue: Make scene.create / scene-mutation benchmarking comparable through a plugin-backed editor lane or an explicitly shared bridge path

Summary

The current CLI-vs-MCP validation work cannot produce a fair result for scene.create because the compact MCP alias is not runnable in the same headless path used by the CLI prototype. The benchmark artifacts currently classify scene creation as non-comparable, which means the repository still lacks a trustworthy answer for one of the most

Read the thread · 2026-03-12 · closed · 1 comment

Harden the CLI-vs-MCP benchmark harness for reproducible evidence and publication-ready outputs

Issue 1 — Harden the CLI-vs-MCP benchmark harness for reproducibility and evidence capture

Summary

The current GoPeak CLI-vs-MCP validation work is good enough to support a hybrid product recommendation, but the benchmark harness still needs hardening before the results can serve as a stable long-term comparison framework. We should turn the current ad hoc benchmark scripts, smoke runs, normalized evidence, and repo-local artifacts into a reproducible benchmark lane with explicit environm

Read the thread · 2026-03-12 · closed · 1 comment

tool.catalog returns empty results unless tools/list was called first

Summary

tool.catalog depends on cachedToolDefinitions, but that cache is populated only when tools/list is called first.

Result: in a fresh session, calling tool.catalog first returns empty catalog data (totalTools: 0) even though tools are available.

Reproduction

  1. Start server.
  2. Initialize MCP.
  3. Call tool.catalog before tools/list:
{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"tool.catalog","arguments":{"query":"scene"}}}

Observed outp

Read the thread · 2026-03-03 · closed · 1 comment

The remaining reports are on the project's issue tracker.