Solo dev builds open-source multi-agent mesh where Claude, GPT and Gemini share one room

Radiant_Bug9778 · reddit · 2026-10-13

A solo developer is building an open-source agent mesh that lets Claude, GPT, and Gemini agents collaborate in shared "rooms" with persistent memory, currently for personal use with plans to release publicly — and is crowdsourcing feature ideas.

Notable design decisions already implemented:

Rooms, not chats: group and 1:1 rooms synced across machines; multi-model agents message each other directly, hand off tasks, and review each other's work
Per-agent Linux containers: each agent gets its own Docker/Podman container with its own project directory and persistent conversation
Typed envelopes over a UDS-first protocol: Rust control plane, Tauri v2 + React UI
Decision Inbox: one triaged approval queue across rooms with source lineage, audit trails, and refetching of stale/refused items
Runaway budgets: per-turn caps on tool calls and wall-clock time, with a kill switch for loops
Rot-free memory: append-only bi-temporal store with a librarian agent, provenance-graded trust tiers, and sleep-time consolidation
Trap-tested: deterministic adversarial suites probing prompt injection, fabricated memory, and approval bypass with zero model calls
Roles/floor policies, presence tiles, voice notes with transcription

Original post →

Solo dev builds open-source multi-agent mesh where Claude, GPT and Gemini share one room

More from coding & agent