argon-agents

Argon adapters for AI agent frameworks: sandboxed, versioned MongoDB for LangGraph and Mem0.

Argon versions MongoDB the way Git versions code — branch, time-travel, diff, merge, undo. This package gives agent frameworks the two things plain MongoDB can't:

1. A disposable copy of state to work against. Fork a sandbox with a TTL, point the agent at an ordinary connection string, and production data stays isolated until you explicitly merge the reviewed changes.
2. An adopt-or-reject story for what the agent did. Diff the sandbox, preview a merge and explicitly apply its reviewed plan, undo supported captured ranges within retained history, or let the running API reclaim the sandbox after its TTL.

Install

SDK 0.2.0 requires Argon 2.1.1 or a compatible later 2.1 patch. We recommend Argon 2.1.2: REST branch creation synchronizes the parent's captured writes before forking, imports require an explicitly quiesced source, and Go consumers have valid /v2 module paths. Version 2.1.1 remains compatible; 2.1.0 has a shutdown bug. Release wheels and source archives are available from GitHub Releases.

pip install 'argon-agents @ https://github.com/argon-lab/argon-agents/releases/download/v0.2.0/argon_agents-0.2.0-py3-none-any.whl'
pip install 'argon-agents[langgraph] @ https://github.com/argon-lab/argon-agents/releases/download/v0.2.0/argon_agents-0.2.0-py3-none-any.whl'

These commands install the existing, versioned release wheel. PyPI publication of 0.2.0 is pending publisher configuration; PyPI currently serves 0.1.0. Do not substitute an unpinned PyPI install. The release workflow verifies the registry and installs its wheel in a fresh environment before declaring a publication complete. See release operations.

Requires a running Argon API server (cd api && go run .) backed by MongoDB 7 as a replica set. The managed API waits for capture readiness, synchronizes versioned operations and sweeps expired sandboxes every minute. Stop native writers before release or discard. The public hosted demo does not expose native connection strings.

The client

from argon_agents import ArgonClient

argon = ArgonClient("http://localhost:8080")
argon.get_or_create_project("support-bot")

sandbox = argon.create_sandbox("support-bot", ttl_minutes=60, actor="agent:run-42")
db = sandbox.pymongo_database()        # plain pymongo, isolated copy
db.tickets.insert_one({"_id": "t1", "status": "resolved"})

print(sandbox.diff())                  # what the agent changed
plan = argon.merge_preview("support-bot", sandbox.branch)
print(plan)                            # inspect changes and conflicts
# After reviewing this exact plan, explicitly apply it:
# argon.merge_apply(plan["id"])
# Or reject the proposal with sandbox.discard().

merge_preview creates a plan without changing the target branch. Review that plan before calling merge_apply(plan["id"]); a stale plan must be previewed again. The convenience methods sandbox.merge() and saver.merge() preview and apply immediately, with no pause or approval step. Use them only when that automatic application is intentional in your own workflow.

The actor labels the entire branch/run, not individual MongoDB clients. Use a separate sandbox for each agent. For a protected API, pass ArgonClient(api_url, token=...); never give an agent a production service credential. Inspect argon.capture_status() if capture reports degraded.

New application collections must enable changeStreamPreAndPostImages before rapid updates; the LangGraph and Mem0 adapters do this themselves. An update without exact images or unsupported drop/rename stops capture with an explicit degraded status. Retention limits history unless pinned.

Run the complete review workflow

The two-agent example uses ordinary pymongo and no paid model: both agents start from one pin, propose different order prices, merge the reviewed result, surface the competing conflict, discard it, then undo the adopted change and assert the original data is restored.

ARGON_API_URL=http://localhost:8080 python examples/two_agent_review.py

LangGraph

from argon_agents import ArgonClient, ArgonCheckpointSaver

argon = ArgonClient()
argon.get_or_create_project("support-bot")
saver = ArgonCheckpointSaver.from_sandbox(argon, "support-bot", ttl_minutes=60)

graph = builder.compile(checkpointer=saver)   # any LangGraph graph
graph.invoke(input, {"configurable": {"thread_id": "user-42"}})

plan = argon.merge_preview("support-bot", saver.sandbox.branch)
print(plan)                           # review the checkpoint-store changes
# After review: argon.merge_apply(plan["id"])
# Or saver.discard() to reject, or saver.fork(argon) to try another branch.

ArgonCheckpointSaver is the official langgraph-checkpoint-mongodb saver — same wire format, same semantics — running on an Argon branch. LangGraph's checkpoint ids give step-level rewind within a thread; Argon adds branch-level fork/merge/undo/audit across the whole store.

Mem0

Mem0 speaks MongoDB natively; Argon supplies the versioned sandbox:

Semantic search requires MongoDB Atlas Search or a compatible Search deployment. A plain replica set supports versioned document storage but does not implement $vectorSearch. Provision search indexes separately for each branch; Argon versions documents, not search-index definitions. Configure the LLM/embedder required by Mem0 before running Memory.

from argon_agents import ArgonClient, sandboxed_mem0_config
from mem0 import Memory

argon = ArgonClient()
argon.get_or_create_project("support-bot")
config, sandbox = sandboxed_mem0_config(argon, "support-bot")
memory = Memory.from_config({"vector_store": config})

# ... let the agent read/write memories ...
plan = argon.merge_preview("support-bot", sandbox.branch)
print(plan)                           # review the memory changes
# After review: argon.merge_apply(plan["id"])
# Or sandbox.discard(); the running API also sweeps expired sandboxes.

Reproducible evals: dataset pins

A pin is a named, immutable reference to a branch state. While the pin exists, it protects the history it references from garbage collection and resets. Deleting the pin removes that protection; keep independent backups for loss of the underlying deployment. Pin the eval dataset once, then fork a fresh sandbox from the same retained pin for each run:

argon.create_pin("my-project", "eval-v1", note="golden dataset")

run = argon.sandbox_from_pin("my-project", "eval-v1", ttl_minutes=30)
# ... run the eval against run.connection_string ...
run.discard()          # the pin remains; fork it again while it exists

Tests

pip install -e ".[dev]"
ARGON_REQUIRE_STACK=1 MEM0_TELEMETRY=false pytest

CI checks Python 3.10, 3.12 and 3.14 against Argon v2.1.2 and fails if the stack is unavailable. The dispatch input engine_ref accepts an exact engine commit or release tag; the resolved SHA is logged. Tests exercise mandatory conflicts, actual undo, LangGraph invoke/ainvoke and fork isolation, pinned input, and Mem0's real MongoDB insert/get/update/delete methods. The Mem0 document test bypasses Atlas index creation only; it does not claim semantic-search coverage on plain MongoDB.