Salt MCP
The MCP endpoint of a Salt workspace on your server: an agent creates pages, databases and comments with a person rights.
- 178 ★ on the repository
- Last commit
- AGPL-3.0
Broad access
Why this level
- An agent creates, changes and trashes pages and database rows with your user rights.
- The Salt server is self-hosted on your machine and stores team documents and attachments.
Install
- Works in
Manual install
docker run -d -p 8420:8420 -v salt-data:/data ghcr.io/saltmd/salt.md:latestStarts the Salt server; the MCP client connects to the `/mcp` address. Without Docker use `curl -fsSL https://raw.githubusercontent.com/saltmd/salt.md/main/install.sh | sh`.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- An agent creates, changes and trashes pages and database rows with your user rights.
- The Salt server is self-hosted on your machine and stores team documents and attachments.
This is third-party code. Review the repository files before installing.
What it does
Salt is a team workspace in one binary on your own server: pages, databases and realtime editing. Claude, ChatGPT, Codex, Cursor, Gemini CLI and any other MCP client connect to the `/mcp` address and work in the same space as people, with the same pages, permissions and history. An agent creates, edits, moves, duplicates, trashes and restores pages, creates databases, changes schemas, adds and queries rows and sets up views. Search applies the same permission checks as for a person. There are bulk imports from JSON without the content passing through the agent, comments with a note trail and presence: a page shows which agent is working on it. A list of actions is closed to agents: creating or deleting accounts, issuing tokens, taking or restoring backups, changing instance settings and deciding who is in a workspace. A credential grants access within its person rights. The activity log keeps agent changes apart from yours.
Who it is for. Teams that keep documents and tables on their own server and want to give agents access with clear permission limits.
Good fit when
- An agent should sort launch notes into pages and build a database from them.
- Thousands of rows from JSON must be loaded into a task table without passing through the model.
- The log should show separately what the agent changed and what staff changed.
- One workspace should be open to Claude, Cursor and Codex on the same rights.
Not a fit when
- You need a cloud service with no server of your own.
- An agent would have to manage accounts, tokens or backups, which the server forbids.
- The AGPL-3.0 license does not fit your distribution terms.
Example request
Create a Salt page with launch notes and a task database from them, then find every row with the blocker status.MCP
- Transport
- http
- Authentication
- API key
Limitations
You must deploy the server yourself: with `curl -fsSL https://raw.githubusercontent.com/saltmd/salt.md/main/install.sh | sh` (read the script first) or through Docker `docker run -d -p 8420:8420 -v salt-data:/data ghcr.io/saltmd/salt.md:latest`. Reverse proxies, TLS and backups are covered in the self-hosting guide. The license is AGPL-3.0, so copyleft terms apply when you offer the service to others. Connecting needs a user credential and the agent sees exactly what that person sees. The MCP tool list and agent access setup live in the project wiki.
How to disable. Remove the MCP entry from the client config and revoke the access granted to the agent in Salt. To remove the server stop the container or systemd service and delete the `salt-data` volume.
FAQ
How do I connect an agent?
Set the `/mcp` address of your Salt server and the credential of the person the agent acts for in the MCP client. The tool list is in the wiki at `salt.md/wiki/mcp-tools/`.
What can an agent not do?
Create or delete accounts, issue tokens, take or restore backups, change instance settings and decide who is in a workspace.
How do I run Salt?
With `curl -fsSL https://raw.githubusercontent.com/saltmd/salt.md/main/install.sh | sh` or Docker `docker run -d -p 8420:8420 -v salt-data:/data ghcr.io/saltmd/salt.md:latest`. No separate database is needed.
Can I see what an agent does?
Yes. The activity log keeps agent changes apart and a page shows presence: which agent is editing it now.
How do I load a lot of data?
Through a bulk import from any JSON source: the content does not pass through the agent.
Discussion
No questions or reviews yet.
Related
MCP serversAn open-source workspace with pages, databases and kanban where Claude and Cursor read and write through MCP
A local MCP server for Mac: Claude, Codex and ChatGPT work in apps, files, terminal and browser under your permissions
A team wiki living in Git with a 51-tool MCP server: agents read and edit pages, tables and boards.
The official Kibela MCP server searches, reads and edits notes, folders and comments of your wiki straight from an assistant chat
An MCP server that lets an agent work on your computer: terminal and long-running processes, file search and edits, Excel, PDF and DOCX
An MCP server with a pet for an agent terminal: it levels up, reacts to your work and in Guard mode flags shaky assumptions
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Salt (this card) | no | API key | 178 | |
| Notion MCP Server | yes | API key | 4,704 | an MCP server for cloud Notion; here your own workspace in one binary with agent deny lists |
| MCP Outline Server: your Outline wiki for agents | no | API key | 158 | connects an agent to Outline; here Salt with bulk JSON import and agent presence |
README in short
It describes a one-command or Docker install, people and agents sharing a workspace, agent access limits and the workspace features themselves: blocks, databases, views, search and collaborative editing. The AGPL-3.0 license is noted separately.