The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
Salt.Md
The open-source workspace for people and AI agents. Docs, databases and realtime collaboration with an MCP server in the same binary, so agents work in the same workspace on…
Install
$
docker run -d -p 8420:8420 -v salt-data:/data ghcr.io/saltmd/salt.md:latestClaude Code
claude mcp add salt-md -- docker run -d -p 8420:8420 -v salt-data:/data ghcr.io/saltmd/salt.md:latestCursor / Claude Desktop (mcp.json)
{
"mcpServers": {
"salt-md": {
"command": "docker",
"args": [
"run",
"-d",
"-p",
"8420:8420",
"-v",
"salt-data:/data",
"ghcr.io/saltmd/salt.md:latest"
]
}
}
}Install command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →Docs, databases and realtime editing for your team. The same workspace is open to Claude, ChatGPT and Cursor, on exactly the access you grant them. One binary, on your own server.
Website · Live demo · Documentation · Quickstart
Quickstart
curl -fsSL https://raw.githubusercontent.com/saltmd/salt.md/main/install.sh | sh
One command downloads the binary for your platform, installs it, starts it and
prints the address to open. On a Linux server it sets itself up as a systemd
service, and running it again upgrades it. No curl on the machine?
wget -qO- … | sh does the same.
Or with Docker:
docker run -d -p 8420:8420 -v salt-data:/data ghcr.io/saltmd/salt.md:latest
There is nothing else to install: no database server, no cache, no object store. Reverse proxies, TLS and backups are in the self-hosting guide.
People and agents in the same place
An agent is asked to organise the launch notes. A page appears, a database is created, the rows fill in. A person opens the same board a second later and carries on. Same pages, same permissions, same history.
Connect Claude, ChatGPT, Codex, Cursor, Gemini CLI or any other MCP client to
/mcp and it works on the workspace you use:
- Pages: create, edit, move, duplicate, trash and restore
- Databases: create them, change the schema, add and query rows, set up views
- Search with the same permission checks a person gets
- Bulk imports from any JSON source, without the content passing through the agent
- Comments, and an append-only note trail on every page
- Presence: the page shows which agent is working on it, before you start typing
And a list it cannot touch. No agent can create or delete accounts, issue tokens, take or restore a backup, change instance settings or decide who is in a workspace, and the server tells every agent that connects. A credential carries its person's access and never more, every workspace decides what agents may do there, and the activity log keeps their changes apart from yours.
MCP tools · Agent access · Permissions
A whole workspace
Similar plugs
Reverse engineer anything with agents, from app behavior down to native binaries.
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype…
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
Open-source web crawler and scraper for LLMs and AI agents: any website into clean, LLM-ready Markdown. Run it yourself, or use Crawl4AI Cloud with one key.