# ctxora

Source: https://glimind.com/server/npm%2Fctxora
Registry id: `npm/ctxora`
Status: unknown

## Install

```json
{
  "mcpServers": {
    "ctxora": {
      "command": "npx",
      "args": [
        "-y",
        "ctxora"
      ]
    }
  }
}
```

Claude Code:

```sh
claude mcp add ctxora -- npx -y ctxora
```

## Tools (29)

- `context_stats`: Return workspace-scoped index, graph, bundle and token statistics.
- `delete_conversation_handoff`: Delete one handoff without affecting other workspaces.
- `ecc_search`
- `ecc_status`
- `estimate_tokens`: Estimate token count for any text. Useful for checking prompt size before sending to model.
- `get_token_budget`: Return the dynamic token budget for a specific model. Shows context window, reserved headroom, and available inject budget.
- `handoff_conversation`: Prepare an automatic no-compression handoff when history is too long. The original provider-format JSON is stored unchanged only after the token threshold is re
- `invalidate_cache`: Clear the retrieval cache. Call when workspace files have changed.
- `invalidate_context`: Invalidate one workspace index or bundle store.
- `list_conversation_handoffs`: List auditable handoffs for one workspace.
- `memory_delete`: Delete a memory entry by exact key and type.
- `memory_evict`: Evict least-recently-used memory entries to free space.
- `memory_inject`
- `memory_list`: List recent memory entries, optionally filtered by type.
- `memory_save`: Save knowledge to long-term memory. :param key: Short identifying key (used for search). :param value: Content to remember. :param mtype: Memory tier — "episodi
- `memory_search`: Search long-term memory by semantic similarity. :param query: Search query. :param mtype: Filter by tier: "episodic" | "semantic" | "procedural" | "" (all) :par
- `memory_stats`: Return statistics per memory tier (count, tokens, hits).
- `plan_context`: Choose an explainable, overridable context strategy.
- `prepare_context`: Execute the selected CAG, RAG, long-context or graph strategy.
- `purge_expired_handoffs`: Purge handoffs past their configured retention window.
- `refresh_workspace`: Incrementally refresh authorized chunks, embeddings and code graph.
- `register_workspace`: Register canonical allowed roots and indexing limits.
- `reindex_paths`: Force re-chunking and re-indexing of given paths. Use after significant file edits to refresh the vector index.
- `restore_conversation_handoff`: Restore an explicitly requested, unmodified conversation handoff.
- `retrieve_context`: Return structured provider-neutral evidence and coverage diagnostics.
- `retrieve_context_legacy`
- `route_skills`: Automatically route a non-trivial task to relevant installed skills and retrieve guidelines. Must be invoked for non-trivial tasks before implementation.
- `skill_feedback`: Submit post-task evaluation and skill feedback (success, failure, corrected) after validation and self-evaluation using the retained route_id.
- `skill_learning_status`: Check skill learning statistics and reranking status.

Measured by Glimind (https://glimind.com): MCP handshake + tool list only, no tool is ever called. CC BY 4.0.