AI_APICallError: GenerateContentRequest.tools[0].function_declarations[32].parameters. required[0]: property is not defined with Gemini API #3140

Activity

1. github-actions commented on Oct 13, 2025

   

   github-actionsBot

   on Oct 13, 2025 – with GitHub Actions

   Contributor

   More actions

   This issue might be a duplicate of existing issues. Please check:

   • [bug] google provider not running due to tool call schema mismatches #365: [bug] google provider not running due to tool call schema mismatches - Both issues report the same type of Gemini API error with function_declarations schema validation failing, specifically around 'parameters.required' properties not being defined

   Feel free to ignore if none of these address your specific case.
2. rekram1-node commented on Oct 13, 2025

   

   rekram1-node

   on Oct 13, 2025

   Contributor

   More actions

   Yeah this is due to mcp server tools, we can sanitize them but they aren't compatible with gemini api out of the box
3. rekram1-node

   self-assigned this

   on Oct 13, 2025
4. rekram1-node commented on Dec 1, 2025

   

   rekram1-node

   on Dec 1, 2025

   Contributor

   More actions

   This should fix it: 0331931

   in next release
5. rekram1-node

   closed this as completedon Dec 1, 2025
6. jeffscottward commented on Dec 18, 2025

   

   jeffscottward

   on Dec 18, 2025

   Last edited by jeffscottward

   More actions

   I am still getting this error.
   I am trying to use Docker MCP toolkit as an aggregation for all my MCPs to stop it from eating the context window.
   My current open code version is 1.0.167
   @rekram1-node should this be reopened?

   Specifically, my error says exactly this:

   * GenerateContentRequest.tools[0].function_declarations[21].parameters.required[1]: property is not defined* 

   Since I think I'm already at the latest version, and that this shouldn't be an issue, I asked AI, and it said this.

   ---

   If you're already on a version that includes commit 0331931, then the issue is likely one of these scenarios:

   Possible Explanations

   1. The Fix Doesn't Cover All MCP Servers
      The commit may have fixed specific servers mentioned in issue AI_APICallError: GenerateContentRequest.tools[0].function_declarations[32].parameters. required[0]: property is not defined with Gemini API #3140, but the Docker MCP Gateway itself might be generating malformed schemas for other MCP servers (like the ones you're using). The fix in 0331931 may have been targeted at OpenCode's built-in tool declarations, not the MCP Gateway's schema translation layer.

   2. Docker MCP Gateway Issue (Not OpenCode)
      The problem might actually be in Docker Desktop's MCP Gateway itself, not OpenCode. The Gateway is responsible for translating MCP server schemas into a format that AI models can consume. If Docker's Gateway has bugs in how it generates Gemini-compatible schemas, OpenCode can't fix that—Docker would need to release an update to their MCP Toolkit.​

   3. Specific MCP Servers Have Invalid Schemas
      The MCP servers you enabled (GitHub, Neon, Next.js, etc.) might have invalid tool declarations in their own source code. When the Gateway passes these through to OpenCode and then to Gemini, the validation fails. This would require fixes from the individual MCP server maintainers.

   What You Can Check
   To determine where the issue actually is:

   1. Check Your OpenCode Version
      Run this command to see exactly which version you're on:

   bash
   opencode --version
   The fix from commit 0331931 should be in version 1.0.137 or later (based on the Chocolatey package dated December 17, 2025).​

   2. Check Docker MCP Toolkit Version
      The issue might be in Docker's MCP Gateway, not OpenCode. Check your Docker Desktop version:

   Open Docker Desktop

   Go to Settings → General

   Look for "MCP Toolkit" or check the Docker version

   3. Test Without Docker MCP Gateway
      Temporarily remove the MCP_DOCKER configuration from your OpenCode config to see if errors persist with other tools. If errors disappear, it confirms the Gateway is the problem.

   4. Check Which Specific Tools Are Failing
      The error indices (13, 63, 103, 114) point to specific tools. You can inspect the actual schema being sent to Gemini by enabling debug logging in OpenCode or checking the Gateway's output directly.

   Most Likely Scenario
   Given that you're on the latest OpenCode version and the issue persists, this is almost certainly a Docker MCP Gateway bug, not an OpenCode bug. The Gateway is translating MCP server schemas incorrectly for Gemini's strict validation requirements.

   Your current workaround (disabling problematic MCP servers) is the right approach until Docker releases an updated MCP Toolkit that properly formats function declarations for Gemini's API. Other AI models like Claude are more lenient with schema validation, which is why this specifically affects Gemini users.​

   👍React with 👍1Reacted by BlueT - Matthew Lien - 練喆明
7. fwkyle

   added a commit that references this issue on Feb 3, 2026

   fix(provider): strip properties/required from non-object types in Gem…

   ...

   cb71141