BACK TO MCP SERVERS
dataany

Argon

MongoDB sandboxes for AI agents, Git for MongoDB underneath: branch, diff, merge, undo over MCP.

At a glance

Not tested yet
GitHub stars
179
Weekly downloads
2,905
Last commit
2026-11-19
Licence
MIT
Transport
stdio
Auth
none
Install
npx -y argonctl
Stats checked 2026-11-19

What it does

Argon gives each AI agent run its own MongoDB database, forked from your data. An agent connects to the sandbox with any MongoDB driver, and the writes it makes are captured as a diff. You can then review the diff, merge it as a data pull request, or discard it. Underneath, it works like Git for MongoDB: branch, diff, merge, time travel and undo. It is for people who want agents to work against real MongoDB data without touching the source until they approve the changes.

Argon, MongoDB sandboxes for AI agents. Git for MongoDB underneath: branch, diff, merge, undo.
Image: argon-lab/argon (MIT)

Tools it gives your agent

The MCP server exposes 13 tools covering sandbox, diff, merge, undo and pins. The sources name only one tool directly:

  • argon_merge_apply: applies a merge plan. You can withhold it with argon mcp --disable-tools argon_merge_apply to keep a person in the loop.

Quick start

brew install argon-lab/tap/argonctl
npm install -g argonctl --allow-scripts=argonctl
docker run -d --name argon-mongo -p 127.0.0.1:27017:27017 mongo:7 --replSet rs0
docker exec argon-mongo mongosh --quiet --eval 'rs.initiate({_id:"rs0",members:[{_id:0,host:"localhost:27017"}]})'
argon doctor
argon import database --uri mongodb://localhost:27017 --database myapp --project myapp --source-quiesced
argon sandbox create -p myapp --ttl 1h --watch --actor agent:run-1
argon diff          -p myapp -b <sandbox>
argon merge preview -p myapp -b <sandbox>
argon merge apply <plan-id>
argon sandbox discard -p myapp -b <sandbox>
claude mcp add argon -- argon mcp

MongoDB must run as a replica set; a one-node set is fine. For npm, use Node.js 24 LTS with npm 11.16 or newer. Point the agent at the connection string the sandbox prints.

Add it to Claude Code from a terminal:

claude mcp add argon -- npx -y argonctl

Or add this to .mcp.json in your project root (Claude Code) or .cursor/mcp.json (Cursor):

{
  "mcpServers": {
    "argon": {
      "command": "npx",
      "args": [
        "-y",
        "argonctl"
      ]
    }
  }
}

Environment variables

  • ARGON_SOURCE_URI: the source database URI, read per command and never stored.
  • ARGON_EXPERIMENTAL_SOURCE: set to 1 to enable the experimental options --auto-resync and --enable-source-images.

Example prompts

  • Create a sandbox from my project myapp with a one hour TTL and give me its connection string.
  • Show me the diff of what the agent changed in this sandbox.
  • Preview a merge plan for this sandbox so I can review it before applying.
  • Discard this sandbox and throw away the run's changes.

What to watch out for

  • MongoDB must run as a replica set.
  • Write-back tests have not covered Atlas, deployments with access control, multi-member elections, sharded sources, or MongoDB versions other than 7.0.
  • Write-back and --enable-source-images require importing with a user allowed to run replSetGetConfig, such as one with the clusterMonitor role.
  • The MCP server and REST API do not offer write-back.
  • Each project can run one follower or resync at a time.
  • Pins protect captured document state while they exist; keep independent backups.
  • saver.merge() previews and applies immediately; use separate calls when approval is required.

Not tested by tools4agents yet. Read the source repository before you install it.

GitHub SourceGitHub Stars
Last verified 2026-11-19

Similar tools