FAQ

Frequently asked questions

Short answers about running agents against MongoDB. Ask the maintainers in Discussions.

01What is Argon?

Argon is an MIT-licensed, self-hosted versioning layer for MongoDB. It adds branching, time travel, reviewed merge and per-actor undo. The hosted console is a temporary sample-data demo. Explore the features.

02How is a branch different from a database dump?

A dump is a standalone copy without Argon's branch history or merge workflow. A branch is lightweight metadata. Checkout materializes a real MongoDB database; readiness, disk use and first-query costs depend on the dataset. How database branching works.

03Can I use my existing driver?

Yes. argon checkout materializes a branch and prints a MongoDB connection string. Drivers, mongosh and Compass connect to that database. Keep capture healthy while writing, and prepare new collections for exact images before updates. Start locally.

04What are the requirements?

Use a MongoDB replica set, including a single-node local set, or a compatible Atlas deployment. Change streams do not work on a standalone mongod. Exact pre/post images and the necessary MongoDB permissions are required for supported restoration. Run argon doctor before a workload. Set up MongoDB.

05Are branches an access-control boundary?

Experiments write to a separate branch database. Applying a reviewed merge explicitly changes the target branch. Use MongoDB credentials and network controls for access isolation. Operating limits.

06Can every change be undone?

A healthy capture process records document changes. Undo requires complete images and retained history. Missing images or unsupported drop/rename operations mark history incomplete and prevent unsafe restoration. Native driver writes use the actor configured for the branch or run. Separate agents need separate branches; Argon does not identify individual clients sharing one connection. Capture and attribution.

07How far back can I time-travel?

Time travel is available within retained history. Pins protect the data they reference while the pin exists; keep backups and choose retention for your workload. Time travel and backups.

08Who manages capture and sandbox expiry?

REST and MCP manage capture and TTL cleanup while the service runs. Standalone CLI sandbox creation requires a running watch process and scheduled sandbox sweep. Connect an agent.

09How fast is branching?

Creating branch metadata does not copy documents. Physical checkout, sandbox readiness, historical queries and MongoDB storage have separate costs that depend on the dataset and workload. Benchmark numbers apply to the documented engine version and measurement scope. Reproduce the benchmarks.

10How do I install Argon?

Engine 2.1.2: npm install -g argonctl@2.1.2. Python SDK 0.2.0: python3 -m pip install "argon-agents[langgraph]==0.2.0". The SDK command installs the matching Git release; use a Python virtual environment. Run the full setup.