mlorentedev/hive

[![hive MCP server](https://glama.ai/mcp/servers/mlorentedev/hive/badges/score.svg)](https://glama.ai/mcp/servers/mlorentedev/hive) 🐍 🏠 🍎 πŸͺŸ 🐧 - On-demand Obsidian vault access via MCP. Adaptive context loading (67-82% token savings), full-text and ranked search, health checks, auto git commit, and worker delegation to cheaper models. 10 tools, works with any MCP client.

VERIFIED 10 toolsΒ·uvxΒ·v3.2.4
UVXNO AUTHMIXEDSETUP LOWSTARTS CLEANSCHEMA LOW10 TOOLS

Install

uvx hive-vault

Capabilities

Server instructions

Hive provides on-demand access to an Obsidian vault. ## When to use each tool - **Start of session:** Call `session_briefing(project)` to load tasks, lessons, git activity, and health in one call. - **Browsing vault structure:** Use `vault_list` to discover projects and navigate directories. Call `vault_query` to read their content. - **Reading vault files:** Use `vault_query` instead of direct filesystem access. Supports section shortcuts (context, tasks, roadmap, lessons) and arbitrary paths. Use `scope:project` syntax when project names are ambiguous across scopes. - **Finding information:** Use `vault_search` for keyword/regex lookup with filters. Add `ranked=True` for relevance-ranked results, or `since_days=N` for recent changes. - **Recording lessons:** Call `capture_lesson` when a bug fix, architectural decision, or useful insight emerges during work. Use `capture_lesson(text=...)` for bulk extraction from large text blocks. - **Writing to vault:** Use `vault_write` to create, append, or replace files. Use `vault_patch` for surgical find-and-replace edits. - **Offloading work:** Use `delegate_task` for summarization, boilerplate generation, or analysis. Use `delegate_task(project=...)` to summarize vault files. Best for boilerplate and format conversion, not for architecture or security decisions. - **Checking workers:** Call `worker_status` before `delegate_task` to verify budget and model availability. - **Checking vault health:** Call `vault_health` after modifying vault files or periodically to detect drift. Add `include_usage=True` for usage statistics. Use `project='_meta'` to access cross-project content (patterns, templates). Read-only tools are safe to call freely. Write tools auto-commit to git.