Unified Environment for Six Computers: An Alternative to Claude Code Team Plans

Organizing efficient teamwork with AI agents often runs into the limitations of team pricing tiers and context loss between sessions. A developer shared a solution that connects six heterogeneous machines into a unified workspace while preserving continuous context.

Infrastructure and Interaction Logic

The system connects six devices across three team members:

• an always-on desktop (acting as the primary hub);
• two Windows laptops;
• two macOS computers;
• a remote Linux VPS.

Each computer runs its own Claude Code account. At the same time, any team member can pick up a task started by a colleague on any machine without re-explaining the context in chats—the AI agent accesses the up-to-date context of the distributed system.

Synchronization and Coordination

Connecting the devices does not rely on centralized cloud databases or dedicated servers. The entire architecture is built on simple local tools:

• Syncthing acts as the transport layer, synchronizing isolated directories: scripts, knowledge bases, configurations, and working memory.
• Two-way exchange: interaction between machines occurs via a file-based task queue and is mirrored to a private Telegram group for status tracking.
• File locks: before editing a shared document, the agent acquires a 15-minute temporary lock to prevent concurrent write conflicts.
• Cross-platform compatibility: scripts contain no hardcoded paths; parameters are read from the local configuration file of each specific OS.
• Security: credentials and secret keys are isolated in a separate directory and excluded from the shared knowledge base.

Context Management via Obsidian

The knowledge base and guidelines are stored in a shared Obsidian vault. To avoid overloading the initial context and to manage token consumption, the knowledge base is tiered:

• Main rules file (CLAUDE.md): contains brief triggers and pointers to detailed instructions. It is read at the start of each session. Editing permissions for this file are restricted to a single person.
• Detailed guidelines: stored as separate notes in the Obsidian vault and loaded by the agent only on demand.

According to the author's benchmarks, the median initial session context size is around 91,500 tokens. Offloading full rule texts into on-demand notes helps conserve the context window.

Key Advantages of the Approach

1. Seamless context: task handoff happens automatically, eliminating the need to manually sync session outputs.
2. Fast node onboarding: adding a new computer only requires copying the configuration and starting folder synchronization.
3. Subscription cost savings: each member uses an individual account, avoiding overpaying for team licenses with high minimum seat counts.
4. Vendor independence: rules are formatted as plain text files, making it easy to integrate other AI agents without overhauling the architecture.