token-context-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| TOKEN_CONTEXT_CONFIG | No | Optional path to an explicit shared/portable TOML configuration file for the repository registry. By default, the registry is stored globally for the current user at %APPDATA%\token-context-mcp\repos.toml. |
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 |
|---|---|
| list_repositoriesA | List registered repository IDs. Call this first; roots are never exposed. |
| get_repo_mapB | Ranked definitions for a repo_id within budget. Compact entries are [id, path:line, kind/name]; format='full' adds signatures and evidence. For orientation, not proof of full coverage. |
| find_symbolsC | Find source-backed symbols by name or qualified-name pattern. Returns IDs and spans, never arbitrary files. |
| get_module_dependentsC | Lexical import relationships for indexed path or module. Dynamic imports flagged. Not semantic resolution. |
| search_sourceB | Full-text search ranked by enclosing symbol. expand='graph' (default for profile 'locate') adds resolved callers/callees as neighbors. Call before reading files. |
| get_file_skeletonC | Imports and source-backed headers from one file. Function bodies elided by default. |
| get_symbol_contextC | Bounded source packet for indexed symbol and observed edges. Check freshness and ambiguity warnings. A body larger than max_tokens is read in windows with body_offset_line. |
| get_impact_sliceC | Traverse observed caller/callee edges from a symbol. Candidate impact slice, not complete blast radius. |
| get_index_statusB | Active snapshot metadata and paths changed since indexing. Run before relying on graph results. |
| inspect_symbolC | Resolve a symbol and its immediate graph in one call. view='full' returns a context packet: body, callee/caller signatures, related tests. Use before editing a symbol. |
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 10 tools
Most tools target distinct phases (list repos, index status, search, orient, read context, traverse graph), but boundaries blur in places: get_symbol_context and inspect_symbol both return source packets plus edges for a symbol, and get_impact_slice, get_module_dependents, and search_source(expand='graph') all traverse relationships. Descriptions attempt to differentiate via 'one call', 'lexical vs semantic', and 'candidate vs complete', but an agent could still misselect among them.
Every tool follows a consistent verb_noun snake_case pattern (get_*, find_*, list_*, search_*, inspect_*). Prefixes map cleanly to intent: get for retrieval, find/list/search for discovery. No mixed conventions.
Ten tools is well-scoped for a code-context/index server. Each tool earns its place across discovery, search, orientation, symbol reading, and graph traversal, with no redundant filler.
Coverage spans repository listing, index freshness, symbol search, orientation maps, file skeletons, bounded source context, and impact traversal — a solid read-only surface with no obvious dead ends. Minor gaps like direct line-range/file-read or cross-repo comparison are workable around via existing tools.