Knowledge management MCP server for AI projects - search, tag, version, and organize your project knowledge
$git clone https://github.com/labyrinth-analytics/loredocsInstalls into the current project.
Install loredocs by running `git clone https://github.com/labyrinth-analytics/loredocs`, then use it for the current task and follow its documentation at https://github.com/labyrinth-analytics/loredocs.
| 1 | # LoreDocs v0.1.14 |
| 2 | |
| 3 | Your AI project's knowledge base. Organized, searchable, version-tracked. |
| 4 | |
| 5 | LoreDocs gives Claude persistent access to your project documentation -- specs, guides, architecture decisions, reference docs -- so it never loses context between sessions. Works with Claude Code, Cursor, OpenAI Codex, and Hermes Agent. |
| 6 | |
| 7 | > **Available on the Anthropic Marketplace.** Install directly from Claude, or via PyPI: `uvx loredocs` |
| 8 | |
| 9 | ## Quick Start |
| 10 | |
| 11 | **Prerequisites:** [uv](https://docs.astral.sh/uv/getting-started/installation/) (fast Python package manager). |
| 12 | |
| 13 | ```bash |
| 14 | # Install uv (one time) |
| 15 | curl -LsSf https://astral.sh/uv/install.sh | sh |
| 16 | |
| 17 | # Clone and install |
| 18 | cd /path/to/loredocs |
| 19 | uv sync |
| 20 | ``` |
| 21 | |
| 22 | For detailed installation instructions, see [INSTALL.md](INSTALL.md). |
| 23 | |
| 24 | ## Using LoreDocs |
| 25 | |
| 26 | ### Claude Code (Terminal) |
| 27 | |
| 28 | ```bash |
| 29 | claude --plugin-dir /path/to/loredocs |
| 30 | ``` |
| 31 | |
| 32 | Or inside an existing session: |
| 33 | |
| 34 | ``` |
| 35 | /plugin add /path/to/loredocs |
| 36 | ``` |
| 37 | |
| 38 | Once loaded, Claude has access to all 42 LoreDocs MCP tools automatically. Ask Claude to "create a vault for this project" or "find the architecture doc" and it uses the tools on its own. |
| 39 | |
| 40 | ### Cowork (Desktop App) |
| 41 | |
| 42 | 1. Click **+** next to the prompt box |
| 43 | 2. Select **Plugins** > **Add plugin** |
| 44 | 3. Browse to the `loredocs` source folder |
| 45 | |
| 46 | **Shared Database Access:** Cowork runs in a sandboxed VM. To access docs saved from Claude Code, ask Claude: |
| 47 | |
| 48 | > "Mount my ~/.loredocs folder" |
| 49 | |
| 50 | ## How It Works |
| 51 | |
| 52 | LoreDocs organizes knowledge into **vaults** -- named containers for related documents. Each vault can hold specs, guides, decisions, checklists, or any text you want Claude to remember. |
| 53 | |
| 54 | ``` |
| 55 | ~/.loredocs/loredocs.db <-- SQLite database (metadata, search index) |
| 56 | ~/.loredocs/vaults/<vault-id>/ <-- Document files on disk |
| 57 | ``` |
| 58 | |
| 59 | **Key concepts:** |
| 60 | |
| 61 | - **Vaults** group related docs by project or topic |
| 62 | - **Documents** are text files with metadata (tags, categories, priority, notes) |
| 63 | - **Version history** tracks every change to every document |
| 64 | - **Full-text search** via SQLite FTS5 finds anything instantly |
| 65 | - **Injection** loads vault content into Claude's context on demand |
| 66 | |
| 67 | ## Your Data is Always Available |
| 68 | |
| 69 | LoreDocs works through MCP tools when they are available and falls back to bundled scripts automatically when they are not. Your vault documents are safe regardless of MCP status -- the same add, search, and retrieve operations work either way. You do not need to configure anything; the plugin skill handles the switch silently. |
| 70 | |
| 71 | ## Verify Installation |
| 72 | |
| 73 | After installing, verify LoreDocs is working by asking Claude: |
| 74 | |
| 75 | > "Run `vault_list` and show me the results." |
| 76 | |
| 77 | If you see a list of vaults (or an empty list if this is your first time), LoreDocs is connected. If you get an error about missing tools, re-run `uv sync` and reload the plugin. |
| 78 | |
| 79 | ## Recommended CLAUDE.md Setup |
| 80 | |
| 81 | For the best experience, add the following snippet to your `~/.claude/CLAUDE.md` (global) or your project's `CLAUDE.md`. This tells Claude how to use LoreDocs consistently across sessions. |
| 82 | |
| 83 | ```markdown |
| 84 | ## LoreDocs (persistent project knowledge) |
| 85 | |
| 86 | At session start: |
| 87 | 1. Call `vault_list` to see available knowledge vaults. |
| 88 | 2. Call `vault_inject_summary` for any vaults relevant to the current project. |
| 89 | 3. Use this context to understand project architecture, decisions, and reference docs. |
| 90 | |
| 91 | During the session: |
| 92 | - If you create significant documentation, add it to LoreDocs with `vault_add_doc`. |
| 93 | - Tag documents for easy cross-vault discovery with `vault_tag_doc`. |
| 94 | |
| 95 | At session end: |
| 96 | - If new docs were created or updated, ensure they are stored in LoreDocs for future sessions. |
| 97 | ``` |
| 98 | |
| 99 | **For Cowork users:** Cowork does not run hooks automatically. Add instructions to call `vault_list` and `vault_inject_summary` at session start in your project CLAUDE.md. |
| 100 | |
| 101 | ## Canonical Project Knowledge |
| 102 | |
| 103 | In multi-agent environments, different tools and agents often create improvised mirrors |
| 104 | of shared skill or configuration content -- playbooks, style guides, shared reference |
| 105 | docs. Those mirrors drift. One agent updates the source; the other keeps reading the |
| 106 | stale copy. Two agents in the same project end up operating from divergent knowledge |
| 107 | with no visible signal that anything is wrong. |
| 108 | |
| 109 | LoreDocs prevents this by making the vault the single canonical source that every agent |
| 110 | reads. Instead of each agent loading a local file copy, every agent calls |
| 111 | `vault_inject_by_tag` at session start and gets the same vault-managed version. |
| 112 | |
| 113 | ### Recommended pattern |
| 114 | |
| 115 | Store shared content (playbooks, team guidelines, shared specs) as vault documents |
| 116 | rather than as local files that agents copy or mirror. |
| 117 | |
| 118 | Agents load the content at session start: |
| 119 | |
| 120 | ``` |
| 121 | vault_inject_by_tag: team-playbook |
| 122 | ``` |
| 123 | |
| 124 | All agents -- regardless of surface (Claude Code, Cowork, CLI, or any future AI tool) |
| 125 | -- call the same vault and receive the same current version. Updating the content |
| 126 | requires editing the vault document once; all agents pick up the change on their next |
| 127 | session start. |
| 128 | |
| 129 | Local files (`.claude/skills/`, |