RE: MCP Servers - Potential issue?

[mynameistito]

Servers that were previously working are now returning Operation timed out after 5000ms. This affects local stdio servers (like chrome-devtools, context7, sequential-thinking) and seems to affect the general stability of the MCP client.

Additionally, the github server is returning an SSE error (400), and sentry is reporting a missing OAuth state, suggesting a potential broader issue with state management or process spawning in the new version.

Steps to Reproduce

1. Define multiple MCP servers in opencode.json (mix of local uv python scripts, node scripts, and SSE endpoints).
2. Start a new OpenCode session.
3. Observe the "MCP" section in the sidebar/logs.
4. Wait for initialization.

Expected Behavior

MCP servers should initialize within the default timeout window, or the timeout should be configurable/longer. Servers that were working in the previous version should continue to connect.

Actual Behavior

• Timeouts: Most local servers fail with: Operation timed out after 5000ms.
• GitHub SSE: Fails with SSE error: Non-200 status code (400).
• Sentry: Fails with No OAuth state saved for MCP server: sentry.

Screenshots / Logs

Session Screenshot:

▼ MCP
• chrome-devtools: Operation timed out after 5000ms
• cloudflare-docs: Operation timed out after 5000ms
• cloudflare-observability: Operation timed out after 5000ms
• context7: Operation timed out after 5000ms
• github: SSE error: Non-200 status code (400)
• next-devtools: Operation timed out after 5000ms
• playwriter: Operation timed out after 5000ms
• sequential-thinking: Operation timed out after 5000ms
• sideways: Operation timed out after 5000ms

## **Configuration**
My `opencode.json` configuration can be viewed here:
https://gist.github.com/mynameistito/94d0f59b11bbb45ec3fc1193a30f0c5d

## **Environment**
*   **OS:** Windows 10
*   **OpenCode Version:** v1.1.1

## **Additional Context**
*   This issue appeared I assume following the update to v1.1.1.

[rekram1-node]

I bumped default timeout to 30 seconds instead of 5 but thats pretty crazy it times out for you? You can also overrie the timeout for mcp servers in ur config

[mynameistito]

I bumped default timeout to 30 seconds instead of 5 but thats pretty crazy it times out for you? You can also overrie the timeout for mcp servers in ur config

Yeah no idea why, but it seems to have cooked out on the latest update.

Um how would I go about adding it to my opencode.json? and changing the timings?

Edit: Disregard found it.

[gasatrya]

Terminal inside NeoVim

Standard terminal seems fine (WSL)

[nelwincatalogo]

This is the error i got after debugging:

mcp config:

{
    "shadcn": {
      "type": "local",
      "enabled": true,
      "command": ["npx", "shadcn@latest", "mcp"],
    },
}

OS: Mac

[nelwincatalogo]

This is working now in version 1.1.2 and above

This is the error i got after debugging:

mcp config:

{
"shadcn": {
"type": "local",
"enabled": true,
"command": ["npx", "shadcn@latest", "mcp"],
},
}

OS: Mac