narrow-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| OPENAI_API_KEY | No | OpenAI API key. If set, provider is 'openai' with default model 'gpt-5-nano'. | |
| ANTHROPIC_API_KEY | No | Anthropic API key. If set, provider is 'anthropic' with default model 'claude-haiku-4-5'. | |
| OPENROUTER_API_KEY | No | OpenRouter API key. If set, provider is 'openrouter' with default model 'openrouter/free'. | |
| NARROW_MCP_RIPGREP_PATH | No | Path to the ripgrep executable. | |
| NARROW_MCP_CONTEXT_LINES | No | Number of context lines around matches. | |
| NARROW_MCP_MAX_CANDIDATES | No | Maximum number of candidates passed to the selector. | |
| NARROW_MCP_MAX_JSON_BYTES | No | Maximum JSON file size in bytes to process. | |
| NARROW_MCP_SELECTOR_MODEL | No | Overrides the provider's default model. | |
| NARROW_MCP_SELECTOR_PROVIDER | No | Overrides auto-detection of the LLM provider. | |
| NARROW_MCP_SELECTOR_TIMEOUT_S | No | Timeout in seconds for the selector LLM call. | 3.0 |
| NARROW_MCP_MAX_CANDIDATE_CHARS | No | Maximum number of characters per candidate. | |
| NARROW_MCP_SELECTOR_API_KEY_ENV | No | Overrides which environment variable holds the API key. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| narrow_fileA | Given a path to a large, low-density non-source file (logs, build output, CSV, JSON, HTML) and a specific intent, returns only the verbatim line-numbered spans relevant to that intent, verified against the original file -- never a summary. Refuses source code files; use code-navigation tools for those. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 1 tool
With only one tool, there is no possibility of an agent selecting the wrong operation from this server. The tool's description also clearly delineates supported file types and explicitly excludes source files.
The single name 'narrow_file' follows a clear verb_noun structure and is easy to predict. However, with only one tool there is no larger naming pattern to assess for consistency.
A single tool is at the low end of what feels like a complete server, even though the server is intentionally narrow. The count is not a severe mismatch, but agents have no fallback or related operations available.
For the server's stated narrow purpose, the tool covers the entire workflow: it accepts a qualifying file and intent, returns verified verbatim relevant spans, and handles boundary cases by refusing source files. No obvious missing operation remains within this deliberately small scope.