Sno Station
Local runtime where Codex, Claude Code, Cursor, Hermes Agent and OpenClaw pass tasks over a shared mailbox, hand work to another agent when one hits its usage limit, and review each other's output across vendors, with shared memory. Apache-2.0 licensed.
A local runtime that lets Codex, Claude Code, Cursor, Hermes and OpenClaw share memory, hand off tasks and review each other's work.
What it is
Sno Station is an Apache-2.0 licensed local runtime for running several coding agents as a group. Agents from different vendors exchange tasks through a shared mailbox and draw on shared memory. The README describes three main behaviours: an agent that hits its usage limit can hand unfinished work to another agent, one agent can request a review of another's output from a different vendor, and memory is consolidated overnight by a process the project calls REM. The project states that memory, messages and skills are stored on the user's own computer, with encryption, and that an optional cloud component exists. Installation is described as a single command in the README, which also notes a separate skills repository.
Key features
- Shared mailbox through which agents pass tasks to one another
- Handoff to another agent when one reaches its usage limit
- Cross-vendor review, where the reviewer is a different agent from the author
- Shared memory across sessions, stored locally and encrypted per the README
- Overnight memory consolidation (REM) with proposed skill changes that require user approval
Use cases
- Continuing a long coding task when one agent's quota runs out
- Getting a second opinion from a different model before merging code
- Keeping lessons from past sessions available to all agents in one place
- Reviewing proposed changes to agent skills before they take effect
Who it's for
Developers who already use several coding agents and want them to share context and review each other's work. Teams who want memory kept on their own machines may also find it relevant.
Getting started
The README gives an install command in prose: a single curl-based command run in a terminal, after which the user runs an onboarding step inside a supported agent. Because the exact commands are not reproduced here, consult the project's README for the installation steps.
Alternatives
The project directory lists related entries including DeepResearchAgent, Composio Agent Orchestrator and A2A Protocol. These are related directory entries, not endorsements, and no comparison is implied.
Drafted with AI from the project's public documentation; verify against the source. Written 3 Nov 2026.