Repository navigation
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
🗄️ Data Integrity & Integration | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
Repository: shunkakinoki/dotagents
Length of output: 3003
🏁 Script executed:
Repository: shunkakinoki/dotagents
Length of output: 11066
🌐 Web query:
OpenRouter MCP HTTP transport field type transport mcp.json schema "mcp.openrouter.ai/mcp"💡 Result:
The OpenRouter MCP server, located at https://mcp.openrouter.ai/mcp, is configured as a remote HTTP-based MCP server [1][2]. In the Model Context Protocol (MCP) configuration file (
mcp.json), remote servers do not use atransportfield type; instead, they are identified by the use of theurlproperty [3]. When configuring OpenRouter in yourmcp.jsonfile, you should define it as follows [1][2]: { "mcpServers": { "openrouter": { "url": "https://mcp.openrouter.ai/mcp" } } } Key details regarding this configuration include: 1. Transport Mechanism: While some MCP clients allow you to explicitly define the transport type (e.g., specifyingtype: "http"ortype: "remote"for clarity) [1][4], the core requirement for connecting to OpenRouter is the presence of theurlfield pointing to the specified endpoint [3]. 2. Authentication: The server uses an OAuth-based flow [1][5]. When an unauthenticated request is made to the endpoint, the server returns a401status code, which triggers the client to initiate the OpenRouter OAuth authorization flow [1][5]. Once authorized, the client receives an API key that is subsequently used for authenticated requests [1][5]. 3. Configuration Flexibility: Depending on the specific MCP client (such as Claude Desktop, Cursor, or custom implementations), you may place this configuration in your globalmcp.jsonfile or add it via the client's built-in connector interface [1][2][3]. If your client specifically requires atypefield to distinguish between local (stdio) and remote (http) servers,httpis the correct designation [4][6].Citations:
🌐 Web query:
Claude Desktop mcp.json remote HTTP transport type http url💡 Result:
Claude Desktop does not natively support the HTTP transport configuration (
"type": "http"with a"url") in itsclaude_desktop_config.jsonfile [1][2]. If you attempt to use this configuration, Claude Desktop will likely ignore the entry, fail to load it, or in some cases, silently strip the server configuration from your file [3][1][2]. To connect to remote HTTP-based MCP servers in Claude Desktop, you should use one of the following two approaches: 1. Custom Connectors (Recommended for public services): For remote, internet-accessible HTTP MCP servers (typically with OAuth or API key authentication), use the Custom Connectors feature in Claude Desktop settings [1][4]. This allows Claude's cloud infrastructure to connect directly to your remote server [4]. 2. Local Stdio-to-HTTP Bridge (Recommended for local HTTP servers): If you must connect to a local HTTP server (e.g.,http://127.0.0.1:8080/mcp), you can use a stdio-based proxy such asmcp-remote[1][2]. In this setup, you configure Claude Desktop to runmcp-remoteas a standard stdio process, and the proxy handles the HTTP communication with your server [1][2]. Example configuration for using the stdio-to-HTTP bridge inclaude_desktop_config.json: { "mcpServers": { "my-local-http-server": { "command": "npx", "args": [ "-y", "mcp-remote", "http://127.0.0.1:8080/mcp", "--transport", "http-only" ] } } } Note: Thetype: "http"configuration format is supported by other MCP clients, such as Claude Code (via.mcp.jsonor~/.claude.json), but it is distinct from Claude Desktop's local configuration requirements [5][2][6]. Always ensure your MCP server is configured to use the "Streamable HTTP" transport, which is the current standard for HTTP-based MCP communication [7][8][9].Citations:
🌐 Web query:
Cursor .cursor mcp.json mcpServers transport type url💡 Result:
In Cursor, the configuration of MCP servers is handled within a
mcp.jsonfile, which can be located either at.cursor/mcp.json(for project-specific settings) or~/.cursor/mcp.json(for global settings) [1][2][3]. The configuration schema distinguishes between two primary transport methods: local (stdio) and remote (HTTP-based) [4][5]. Local Servers (stdio) To configure a local server, you define it using thecommandfield [1][2]. Thetypefield is typically "stdio" (though often inferred ifcommandis present) [2][6]. { "mcpServers": { "my-local-server": { "command": "npx", "args": ["-y", "mcp-server"], "env": { "API_KEY": "value" } } } } Remote Servers (URL) For remote servers, you use theurlfield instead ofcommand[1][2][3]. While the official Cursor documentation often demonstrates these without an explicittypefield, some implementations or related client tools usetypeto specify the HTTP transport (e.g., "http" or "sse") [2][5]. { "mcpServers": { "my-remote-server": { "url": "https://mcp.example.com/mcp", "headers": { "Authorization": "Bearer your-token-here" } } } } Key Configuration Notes: - Transport Types: The Model Context Protocol (MCP) officially definesstdioandStreamable HTTPas the current standard transports [7][8][5]. OlderHTTP+SSEconfigurations are considered legacy/deprecated [5]. - Variable Resolution: Cursor supports variable resolution (e.g.,${env:VAR_NAME}or${workspaceFolder}) withincommand,args,env,url, andheadersfields [2][6]. - Priority: If the same server name exists in both project-level and global configuration files, the project-level settings take precedence [1]. - Security: Never hard-code sensitive API keys or secrets directly in your configuration; use environment variables or environment files (envFile) instead [2][3][4].Citations:
🌐 Web query:
Codex /kilo .codex mcp.json mcpServers transport type url💡 Result:
The configuration format for Model Context Protocol (MCP) servers varies depending on the specific tool or environment (e.g., Kilo Code/Kilocode, Codex, or VS Code), though most follow a similar pattern of nesting servers under an mcpServers object within a JSON configuration file [1][2][3][4]. Configuration Summary: 1. Codex: Codex primarily uses a TOML-based configuration file located at ~/.codex/config.toml, where servers are defined under an [mcp_servers.] table [5][6]. However, when using JSON configuration (such as in plugins or project-specific files), it expects the mcpServers wrapper [4]. For URL-based (remote) servers, the transport type is typically identified as streamable_http [7][6]. 2. Kilo Code (Kilocode): Kilo Code configurations are managed in files such as kilo.json or mcp.json [1][3]. The structure uses an mcpServers object where each server defines a type (e.g., local or remote) [1][8]. For remote servers, you must provide a url and optional headers [1][3]. 3. General Structure (.mcp.json): In many implementations, the mcp.json file serves as the standard configuration file [9][10]. The typical structure is: { "mcpServers": { "my-server-name": { "type": "remote", "url": "https://example.com/mcp", "headers": { "Authorization": "Bearer " } } } } Key Transport Considerations: - Transport Types: Common values for the type field include local (or stdio), remote, streamable_http, and sse [1][3][9][10][11]. - URL Requirement: When the type is set to remote, streamable_http, or sse, the url field is strictly required to define the server endpoint [3][6][9][10][11]. - Compatibility: Always check your specific tool's documentation, as some clients (like Claude Code or specific versions of Codex) may strictly require the mcpServers wrapper, while others might support flat configurations [4][12][10][11].
Citations:
🌐 Web query:
"OpenRouter MCP server" "mcp.openrouter.ai/mcp" Claude Code type http💡 Result:
The OpenRouter MCP server, located at https://mcp.openrouter.ai/mcp, is a remote server that allows your AI tools to access live OpenRouter data, such as model catalogs, pricing, benchmark rankings, and your account credits [1][2]. It uses the HTTP transport protocol [1][3]. To connect the OpenRouter MCP server to a client (such as Claude Desktop or other MCP-compatible tools), you generally add the server URL and then perform an OAuth login [1][2]. For example, in command-line environments, you can typically use: claude mcp add --transport http openrouter https://mcp.openrouter.ai/mcp claude mcp login openrouter In JSON configuration files, it is structured as: { "mcpServers": { "openrouter": { "type": "http", "url": "https://mcp.openrouter.ai/mcp" } } } Note that Claude Code itself integrates with OpenRouter primarily by routing the Anthropic API calls to OpenRouter's "Anthropic Skin" (via setting the ANTHROPIC_BASE_URL to https://openrouter.ai/api), rather than connecting to OpenRouter via an MCP server [4]. The OpenRouter MCP server is intended to be used as a tool provider for your agents, allowing them to search for models or check your usage [2][5]. Using the OpenRouter MCP server provides your agent with real-time data to help it decide which model to use, whereas the "Anthropic Skin" configuration allows the agent to actually execute tasks using OpenRouter's infrastructure [2][4].
Citations:
🌐 Web query:
Claude Code MCP configuration transport type http mcpServers type💡 Result:
In Claude Code, when configuring an MCP server that uses HTTP transport, the
typefield in your configuration must be set to"http"[1][2]. While the broader Model Context Protocol (MCP) specification defines a "Streamable HTTP" transport [2], Claude Code specifically requires the literal string"http"for its configuration schema [1][2]. Using other identifiers—such as"streamable-http"—will result in a configuration error [2]. Configuration Details: - Transport Type:"type": "http"[1][3] - Required Field:"url"(the endpoint of your MCP server) [1][4] - Optional Field:"headers"(for authentication tokens or custom headers) [1][5][3] Example JSON Configuration: { "mcpServers": { "your-server-name": { "type": "http", "url": "https://api.example.com/mcp", "headers": { "Authorization": "Bearer YOUR_TOKEN" } } } } You can manage these configurations via the command line usingclaude mcp add --transport http <name> <url>[1][4], or by manually editing your configuration files (.mcp.jsonor~/.claude.json) [1][5]. Note that the older SSE (Server-Sent Events) transport is deprecated in favor of this HTTP transport [1][5].Citations:
Use HTTP-compatible entries for each shared sync target.
.ruler/mcp.json,skills/create-plugin/SKILL.md, and the synced targets do not agree on one HTTP discriminator:transport: "http"is documented, but the synced entries usetype: "http". Codex does not reliably accept thismcp.jsonformat. Either move HTTP MCPs out of these shared configs, switch all targets to a common supported field, or add and validate explicit translation duringmcp-sync.🧰 Tools
🪛 SkillSpector (2.5.1)
[warning] 20: [E1] External Transmission: Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
Remediation: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
(Data Exfiltration (E1))
[warning] 33: [E1] External Transmission: Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
Remediation: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
(Data Exfiltration (E1))
🤖 Prompt for AI Agents