Does the claude-context MCP server auto-sync code indexes after initial indexing?
I'm using the claude-context MCP server (zilliztech/claude-context). After the initial indexing pass, does it auto-sync so only changed code gets re-embedded and updated? Right now I have to manually re-index, and when I do, it says the index already exists and only force=true works, which rebuilds everything from scratch.
Originally asked on GitHub: https://github.com/zilliztech/claude-context/issues/238
6 answers
Auto-sync / incremental index updates are not available as a user-facing feature yet — a maintainer confirmed it is a separate feature request being tracked. The separate stuck-reindex bug is fixed in v0.1.5 (PR #282):
force=truenow clears stale indexing state, and interrupted indexing is automatically reset when the MCP server restarts. If you hit "already being indexed" or a permanently stuck index, update to v0.1.5.What currently exists (confirmed by a user who inspected v0.1.14): the MCP server exposes only
index_codebase,search_code,clear_index, andget_indexing_status— there is no public incremental-sync tool. Internally, incremental refresh does exist (SyncManager.handleSyncIndex()callsContext.reindexByChange(...)), but it is not reachable from an MCP client. WithCLAUDE_CONTEXT_BACKGROUND_SYNC=falseandCLAUDE_CONTEXT_TRIGGER_WATCHER=true, the server does not watch project files directly: it watches the sentinel file at~/.context/.sync-trigger, and an external hook/process must touch that file to request a sync.Practical workarounds until the feature ships: set up an out-of-band hook that touches the sentinel file, rely on background polling, or accept a full
force=truereindex. Ifforce=truestill asks on the latest version, the maintainer asked for the exact error message and MCP server logs and pointed to the project's troubleshooting guide.Did this work for you?What a GitHub user tried on GitHub (Outcome unclear)
Asked whether the tool auto-syncs after initial indexing so only changed code gets re-indexed; answered as not available yet.
Source: https://github.com/zilliztech/claude-context/issues/238 (opens external site)
Did this work for you?What @giancarloerra tried on GitHub (Didn't work)
Tried re-triggering the code index on an existing index; the tool said the index already exists and only offered force re-index, which restarts everything.
Did this work for you?What @dpearson2699 tried on GitHub (Partly worked)
Ran v0.1.14 in a multi-instance stdio setup with CLAUDE_CONTEXT_BACKGROUND_SYNC=false and CLAUDE_CONTEXT_TRIGGER_WATCHER=true; found the server watches the sentinel file ~/.context/.sync-trigger and exposes no public sync tool.
Did this work for you?What @fdterr tried on GitHub (Outcome unclear)
Updated to @zilliz/claude-context-mcp@latest but still got prompted for
force=true; the maintainer asked for the exact error and logs before diagnosing.Did this work for you?What @fdterr tried on GitHub (Outcome unclear)
Reported having to manually update indexes after every change because auto-sync is missing.
Did this work for you?
Related questions
- claude-context: "Error validating collection creation: [object Object]" when indexing codebase
- claude-context: indexing succeeds but Milvus/Zilliz DB stays empty
- How do I stop claude-context from applying .dockerignore when indexing?
- claude-context on Windows: 'Failed to open file' path has no slashes
- Can I use claude-context without an OpenAI key (OpenRouter, Mistral, or local LM Studio)?