RepoChatMCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CHAT_SEARCH_CLAUDE_ROOT | No | Custom root directory for Claude Code session files. | |
| CHAT_SEARCH_CODEX_ARCHIVED | No | Custom directory for Codex archived session files. | |
| CHAT_SEARCH_CODEX_SESSIONS | No | Custom directory for Codex session files. | |
| CHAT_SEARCH_KNOWLEDGE_MODEL | No | Model identifier for HTTP backend (e.g., moonshotai/kimi-k2-0905@groq). | |
| CHAT_SEARCH_KNOWLEDGE_API_KEY | No | API key for HTTP backend. | |
| CHAT_SEARCH_KNOWLEDGE_BACKEND | No | Knowledge indexing backend: off, auto, http, or codex. Default is off. | |
| CHAT_SEARCH_KNOWLEDGE_BASE_URL | No | Base URL for HTTP backend (e.g., https://openrouter.ai/api/v1). | |
| CHAT_SEARCH_KNOWLEDGE_CODEX_BIN | No | Path to Codex binary for codex backend. | |
| CHAT_SEARCH_KNOWLEDGE_MAX_CHARS | No | Maximum characters for chunk splitting (e.g., 500000). | |
| CHAT_SEARCH_KNOWLEDGE_TIMEOUT_MS | No | Timeout in milliseconds for HTTP requests. | |
| CHAT_SEARCH_KNOWLEDGE_HTTP_CONCURRENCY | No | Concurrency limit for HTTP chunk calls (e.g., 3). |
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
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| chat.list_sessionsB | List discovered chat sessions for the current repo. |
| chat.searchC | Search chat messages by text query with relevance scoring. |
| chat.grepC | Grep chat messages by pattern (supports |, &, regex). |
| chat.read_sessionC | Read messages from a specific session. |
| chat.read_linesB | Read specific lines or lines around a keyword/regex from a session. |
| chat.compaction_knowledgeC | Collect live compaction knowledge metadata and return an absolute file path for the compaction snapshot. |
| chat.start_knowledge_indexA | Start repository knowledge indexing as an async job and return immediately with a job ID. |
| chat.get_knowledge_index_statusA | Get the status and progress of an async repository knowledge indexing job. |
| chat.cancel_knowledge_indexA | Request cancellation of an async repository knowledge indexing job. |
| chat.list_knowledge_batchesC | List persisted knowledge batches with canonical combined artifact information. |
| chat.read_latest_knowledgeB | Read paginated text from the latest persisted combined knowledge artifact. |
| chat.read_knowledge_batchB | Read paginated text from a specific persisted knowledge batch artifact. |
| chat.list_knowledge_filesA | List persisted per-chunk knowledge files for a specific batch. |
| chat.read_knowledge_fileB | Read paginated text from a specific persisted per-chunk knowledge file. |
| chat.search_knowledgeB | Search persisted knowledge artifacts directly and return snippet matches. |
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 15 tools
Tools are mostly distinct, clearly separated into session management, chat search, and knowledge indexing. Some overlap exists between chat.search (relevance scoring) and chat.grep (regex), and among the various knowledge reading tools, but descriptions clarify their different purposes.
All tools use snake_case and the chat. prefix, generally following a verb_noun pattern (list_sessions, read_session, start_knowledge_index). There is one deviation ('compaction_knowledge' is noun-like) and a mix of verbs like search/grep/read, but overall the naming is consistent and readable.
15 tools is within the ideal 3-15 range for a focused MCP server. Each tool addresses a specific aspect of chat session access or knowledge indexing, and none feel redundant or unnecessary.
The tool set covers the core lifecycle of knowledge indexing (start, cancel, status, read artifacts) and session discovery/reading. Minor gaps include no ability to list all indexes at a high level (only batches) and no explicit session deletion, but these are not critical for the server's apparent purpose.