VibeDNA Journal
One journal file per project that your AI reads at session start and updates as it works.
Installation
Ajoutez ceci à la configuration MCP de votre client (Claude Desktop : claude_desktop_config.json ; Cursor : .cursor/mcp.json ; VS Code et les autres utilisent la même forme « mcpServers »).
https://github.com/marstudio360/vibedna-journal-mcp/releases/download/v1.0.3/vibedna-journal-mcp-1.0.3.mcpbVariables d'environnement
JOURNAL_ROOTSFolders to scan for *JOURNAL.md files, separated by ;JOURNAL_HOMEWhere the local search index lives. Default ~/.journal
Documentation
Tirée du README du projet sur GitHub, actualisée chaque semaine. Les liens ouvrent GitHub.
VibeDNA Journal MCP
One journal file per project that your AI reads at session start and updates as it works.
Journal gives every project one plain markdown file (*_JOURNAL.md). Your AI reads it when a session starts and appends to it as work ships: decisions with their reasons, bugs with root cause and fix, current state, roadmap, and a handoff note for the next session. Ranked full-text search finds an entry across every project's journal at once. One optional tool adds a rule to your CLAUDE.md so Claude Code keeps the habit without being asked.
- Runs locally over stdio. Python 3.10+ and the
mcppackage. - The journaling core makes no network calls. Your journals stay on your disk.
- Optional marketplace tools browse and download journals from vibedna.ai, only when you call them.
Homepage: https://vibedna.ai/store/journal
Tools (25)
Catch up on a project
| Tool | What it does |
|---|---|
journal_handoff_brief(name) | Short catch-up: header plus the latest entries |
journal_read(name, max_kb) | Full journal contents |
journal_read_section(name, section) | One section, e.g. Current state or Handoff for next session |
journal_list_sections(name) | The H2 sections of a journal |
journal_list(project, last_days) | List the journals in the index |
Log work as it ships
| Tool | What it does |
|---|---|
journal_append(name, entry, tag) | Timestamped entry |
journal_update_section(name, section, new_content, mode) | Replace, append or prepend a section |
journal_create(name, project_dir, body) | New journal with the standard sections |
Search
| Tool | What it does |
|---|---|
journal_search(query, project) | Ranked full-text search (SQLite FTS5) across all journals, with snippets |
journal_recent_changes(last_hours) | Journals updated lately |
journal_stats() | Totals, biggest and freshest journals |
journal_browse(domain, tier, kind, tag) | Browse the local library by metadata |
journal_market_index(kind) | Local feed of journals of one kind, with metadata |
journal_overlap_check(name) | Find journals that overlap with this one |
Setup
| Tool | What it does |
|---|---|
journal_index_refresh() | Rescan the roots for journal files |
journal_config_add_root(path) | Add a folder to scan |
journal_get_claude_rule() | The recommended CLAUDE.md rule text |
journal_install_rule(scope, confirm) | Install that rule (preview unless confirm=True) |
journal_backup_to_drive(name, remote, remote_folder) | Copy journals to your own rclone remote (optional) |
check_for_update() | Compare this copy with the published version |
vibedna_info() | Product metadata |
Marketplace (optional, reaches vibedna.ai)
| Tool | What it does |
|---|---|
journal_market(query, domain) | Browse the VibeDNA journal shelf |
journal_pair(token) | Link this install to a vibedna.ai account |
journal_get(slug) | Download a journal you own, or get a checkout link for one you don't |
journal_library() | List the journals you own |
Install
git clone https://github.com/marstudio360/vibedna-journal-mcp
cd vibedna-journal-mcp
pip install -r requirements.txt
JOURNAL_ROOTS is the list of folders to scan for journal files, separated by ;. Without it, Journal scans ~/Desktop and ~/Documents.
Claude Code
claude mcp add journal --scope user -e JOURNAL_ROOTS="/path/to/your/projects" -- python /absolute/path/to/vibedna-journal-mcp/server.py
Or in a project's .mcp.json:
{
"mcpServers": {
"journal": {
"command": "python",
"args": ["/absolute/path/to/vibedna-journal-mcp/server.py"],
"env": { "JOURNAL_ROOTS": "/path/to/your/projects" }
}
}
}
Then ask Claude: "Install the journal CLAUDE.md rule globally." That is the step that makes it read and write the journals on its own.
Cursor
Add the same mcpServers block to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project).
Claude Desktop
Add the same mcpServers block to claude_desktop_config.json (Settings > Developer > Edit Config), then restart Claude Desktop.
INSTALL.md is a step-by-step guide written so you can paste it into an AI chat and let the AI do the install.
Configuration
| Variable | Default | What it does |
|---|---|---|
JOURNAL_ROOTS | ~/Desktop;~/Documents | Folders to scan for *JOURNAL*.md files |
JOURNAL_HOME | ~/.journal | Where the search index and optional config live |
JOURNAL_TRANSPORT | stdio | http serves streamable HTTP on PORT, for cloud hosting only |
Journal scaffold
journal_create starts every journal with these sections:
## Vision / Why this exists
## Current state
## Roadmap
## Decisions log
## Incidents
## Handoff for next session
## Recent entries
Security
See SECURITY.md for what the server reads, writes and sends.
License
MIT, see LICENSE. Copyright (c) 2026 VibeDNA.
Support: admin@vibedna.ai
Questions fréquentes
Qu'est-ce que VibeDNA Journal ?
One journal file per project that your AI reads at session start and updates as it works.
Comment installer VibeDNA Journal ?
Suivez les étapes d'installation de sa documentation, affichée sur cette page et dans son dépôt.
Faut-il une clé API pour VibeDNA Journal ?
Son entrée au registre n'indique aucune clé ni secret obligatoire. Certaines fonctions peuvent tout de même demander un compte auprès du service auquel il se connecte.
VibeDNA Journal est-il open source ?
Oui. Son code est sur GitHub sous licence MIT.
Qui maintient VibeDNA Journal ?
Il est publié par marstudio360. AI.tn le référence sans le maintenir ni le recommander.
Issu du registre MCP officiel. Être listé n'est pas une recommandation : vérifiez le code et les permissions d'un serveur avant de lui donner accès à vos données.