# MoSearch > MoSearch is an offline-first Windows desktop search tool for local drives and mapped network shares. It includes MoContext, a local Model Context Protocol (MCP) server that lets AI agents retrieve grounded snippets from the MoSearch full-text index, read local files within configured roots, keep lightweight continuity notes, and save/load explicit long-term context digests. ## Primary AI Entry Points - **MoContext MCP endpoint:** `http://127.0.0.1:43210/mcp` - **MoContext REST fallback:** `http://127.0.0.1:43210/v1` - **Scope:** local loopback only; requires MoSearch/MoContext installed, running, and indexed on the user's Windows PC or server. - **MCP discovery:** tools, resources, and resource templates are auto-discovered by MCP clients. Do not hardcode the full tool list when discovery is available. - **Current MCP surface (26.0):** 15 tools — server summary (status, budgets, and local context paths in one call), health, diagnostics, usage guide, indexed file discovery, evidence packs, file reads (whole file, line range, or window), activity notes with recent-activity and briefing views, context-file listing and reading, guarded memory-file writes, and SaveContext/LoadContext digests. - **External field casing:** MoContext uses `snake_case` for MCP tool schemas, REST JSON/query parameters, JSON resources, and support-report JSON. MCP protocol envelope fields keep their spec-defined casing. ## Recommended AI Startup Flow When MoContext tools/resources are available in an AI client: 1. Call `get_server_summary`. One call returns status, budgets and capabilities, and the exact local sys-doc and memory folders, so no separate capability or path call is needed. 2. Call `list_context_files` or read `mocontext://context-files` to discover installed docs and durable memory files. 3. Read `mocontext://docs/ai-usage` or call `get_usage_guide` for deeper tool guidance. If MCP is blocked but local HTTP is allowed, use the REST fallback endpoints: `GET /v1/server-summary`, `GET /v1/capabilities`, `GET /v1/context-paths`, `GET /v1/context-files?include_system_docs=true`, and `GET /v1/usage-guide`. Only call `load_context_digest` / `save_context_digest` when the user types `LoadContext`, `LoadContext `, `SaveContext`, or explicitly enables that behavior. Default `LoadContext` returns budget-capped saved digest records plus a compact working-set hint; treat both as orientation and verify source files before claims or code changes. If `truncated` is true, use omitted counts to narrow the query or fetch a specific digest by id. Treat MoContext activity notes and memory files as hints, not source-of-truth. Retrieve fresh evidence from indexed files before making factual claims, design conclusions, or code changes. ## Public Documentation - [MoContext Overview and Setup](https://www.meauxsoft.com/MoContext.html): general-audience overview and guided client setup, with separate recipes for editors, extensions, and command-line clients. - [MoContext AI Usage Guide](https://www.meauxsoft.com/MoContext_AI_Usage.md): AI-facing MCP workflow guide, retrieval strategies, memory discipline, and budget reference. - [MoContext Human Guide](https://www.meauxsoft.com/MoContext_Human.html): human-facing guide with copyable AI custom-instructions text, one-session prompt text, and a technical appendix. - [MoSearch Change Log](https://www.meauxsoft.com/release-history_MoSearch.json): MoSearch release history. ## Installed Local Files Install paths vary by version. Use the `context_paths` section of `get_server_summary` instead of assuming exact local folders. Mo-Search 26.0 is a per-user install: - **MoSearch root:** `%LOCALAPPDATA%\Programs\Meauxsoft\Mo-Search\` - **MoSearch data/index:** `%LOCALAPPDATA%\Meauxsoft\Mo-Search\` - **MoContext server:** `%LOCALAPPDATA%\Programs\Meauxsoft\Mo-Search\MoContext\MoContext.Server.exe` - **MoContext data:** `%LOCALAPPDATA%\Meauxsoft\MoContext\` (product-owned sys docs and user/AI memory live under this root) Mo-Search 9.50 can still sit side-by-side in Program Files until the user removes it. ## Capabilities - **Local file search:** fast indexing and retrieval of local Windows files. - **Network share support:** indexes mapped network drives configured by the user. - **Grounded AI retrieval:** ranked, budget-capped source/file snippets from the MoSearch full-text index. - **Long-term local context:** installed docs, user/AI-authored Markdown memory files, and SaveContext digest Markdown can be indexed by MoSearch and surfaced through MoContext. - **Offline-first design:** MoContext runs locally; the public website cannot access a user's local MoContext server.