What would you like to be added?
Native support for accessing Model Context Protocol (MCP) resources (e.g., via a read_mcp_resource tool or equivalent capability) and general performance optimizations to reduce latency in interactions involving MCP.
Why is this needed?
Currently, the Gemini CLI lacks native support for interacting with Model Context Protocol (MCP) resources. While it can discover them, the agent does not have tools to list or read them during a conversation. When an agent attempts to access capabilities exposed as resources, it often stalls, enters loops, and falls back to reading help files to understand how to proceed. This leads to significantly high latency (3-9 minutes).
Additional context
Implement native tools in the Gemini CLI to allow the agent to discover and read MCP resources directly, bringing performance parity and feature alignment with industry standards.
- Add list_mcp_resources Tool
Description: Allows the agent to list all available resources exposed by connected MCP servers.
Parameters: None (or optional server name filter).
Purpose: Enables the agent to discover what context is available to it dynamically.
- Add read_mcp_resource Tool
Description: Allows the agent to read the content of a specific MCP resource using its URI.
Parameters: uri (string, required).
Purpose: Provides low-latency, direct access to the data exposed by the resource, preventing documentation-reading fallbacks.
What would you like to be added?
Native support for accessing Model Context Protocol (MCP) resources (e.g., via a read_mcp_resource tool or equivalent capability) and general performance optimizations to reduce latency in interactions involving MCP.
Why is this needed?
Currently, the Gemini CLI lacks native support for interacting with Model Context Protocol (MCP) resources. While it can discover them, the agent does not have tools to list or read them during a conversation. When an agent attempts to access capabilities exposed as resources, it often stalls, enters loops, and falls back to reading help files to understand how to proceed. This leads to significantly high latency (3-9 minutes).
Additional context
Implement native tools in the Gemini CLI to allow the agent to discover and read MCP resources directly, bringing performance parity and feature alignment with industry standards.
Description: Allows the agent to list all available resources exposed by connected MCP servers.
Parameters: None (or optional server name filter).
Purpose: Enables the agent to discover what context is available to it dynamically.
Description: Allows the agent to read the content of a specific MCP resource using its URI.
Parameters: uri (string, required).
Purpose: Provides low-latency, direct access to the data exposed by the resource, preventing documentation-reading fallbacks.