TigrimOS v1.1.0 + Tiger CoWork v0.5.0 — dropped today. Remote agents, swarm-to-swarm, and configurable governance. Self-hosted, free, open source
Open-source platform now lets agent swarms on different machines collaborate like they're on the same box.
The Tigrim project has launched two major updates: TigrimOS v1.1.0 for Mac/Windows with a built-in Ubuntu sandbox, and Tiger CoWork v0.5.0 as a Linux-native server version. Both are free, open-source, and self-hosted, eliminating cloud dependencies. The standout feature is swarm-to-swarm communication, where entire agent clusters on different physical machines can now collaborate as if they were local. This enables practical setups like using a laptop as a coordinator while offloading heavy inference to cloud GPUs or large-scale retrieval to database servers.
Architecturally, the system offers five configurable governance protocols that users can select per job: Star/Hub for deterministic control, Blackboard for distributed auctions, Pipeline for sequential workflows, Mesh for decentralized collaboration, and Bus for parallel processing. Each agent within a swarm can run different LLM backends (Claude, Ollama, GLM, etc.), and sandbox isolation protects the host system. The platform also includes a visual task monitor for debugging agent interactions and supports long-running sessions with checkpoint recovery.
- Swarm-to-swarm communication enables agents on separate machines (laptop + cloud GPU) to collaborate seamlessly
- Five configurable governance protocols (Star, Blackboard, Pipeline, Mesh, Bus) can be selected per task
- Each agent can use different LLM backends (Claude, Ollama, GLM) within the same swarm
Why It Matters
Enables complex multi-agent workflows across distributed hardware without cloud lock-in, giving teams flexible orchestration options.