Exa MCP server returns `title: null` for tools, breaking Dify's MCP integration

Start at the MCP tool conversion path and provider registration flow described in the issue, then reproduce with https://mcp.exa.ai/mcp and Server ID exa-search. Trace how absent title, meta, and outputSchema fields become null and how registration persists rows; done means optional fields remain absent, title falls back to the tool name, failed validation leaves no broken row, and the MCP page can render and delete malformed entries.

Self Checks

• I have read the Contributing Guide and Language Policy.
• This is only for bug report, if you would like to ask a question, please head to Discussions.
• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report, otherwise it will be closed.
• 【中文用户 & Non English User】请使用英语提交，否则会被关闭 ：）
• Please do not modify this template :) and fill in all the required fields.

Dify version

1.17.0 Plugin daemon: langgenius/dify-plugin-daemon:0.6

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

In Dify, add a new HTTP MCP server with URL https://mcp.exa.ai/mcp and Server ID exa-search.
Save.
The MCP server returns this raw response (verified with curl directly against Exa's endpoint):

json
{
"result": {
"tools": [
{
"name": "web_search_exa",
"description": "Search the web for any topic...",
"inputSchema": { ... },
"annotations": { ... },
"execution": { ... }
}
]
},
"jsonrpc": "2.0",
"id": 1
}
Note: no title key at all.

✔️ Expected Behavior

Expected Behavior

Optional fields absent from the MCP server response should remain absent in Dify's internal representation, not be coerced to null.
If title is legitimately null or absent, fall back to the tool's name for display.
Provider registration should be transactional — a validation failure should not persist a broken row.
The Tools → MCP page should always render, even if a provider record is malformed, and allow deletion of the offending entry.
Suggested Fix

In the MCP tool conversion path, replace patterns like:

python
title = tool.get("title") # returns None if absent
with:

python
title = tool.get("title") or {"en_US": tool["name"]}
Or, if the code deliberately distinguishes absent from null, use:

python
title = tool["title"] if "title" in tool else None
and handle None at the point of I18nObject construction rather than serializing null into the intermediate representation.

❌ Actual Behavior

Observed Behavior

Dify's registration fails with:

text
[{"type":"missing","loc":["provider_id"],"msg":"Field required","input":{},"url":"https://errors.pydantic.dev/2.12/v/missing"}]
A row is persisted in tool_mcp_providers, but it contains Dify-injected null fields:

json
{
"name": "web_search_exa",
"title": null,
"meta": null,
"outputSchema": null,
...
}
Dify's client apparently calls tool.get("title"), tool.get("meta"), and tool.get("outputSchema") and serializes missing keys as JSON null — instead of omitting them. The title null then fails I18nObject validation during the provider registration step.