What Happened
Metateam, built by SeriousBit (based in Chișinău, Moldova), launched on Hacker News as a CLI tool that manages multiple AI coding agents inside a single terminal dashboard. It supports Claude Code, Codex CLI, and Gemini CLI running simultaneously through tmux, with a tabbed interface (F1-F11) to switch between live agent sessions.
The tool has been in use internally since October 2025. You install it via a shell script, run metateam in your project directory, and a dashboard launches with a crew coordinator that can summon specialist agents and assign tasks.
Three features stand out beyond basic session management. First, persistent memory: sessions are captured automatically and agents can access relevant context from past work. You can write Knowledge Base entries that persist across sessions, so every future run starts with that context loaded. Second, agent coordination: a crew leader receives a single instruction, delegates to specialist agents, and they communicate in real-time through the dashboard. Third, multi-machine sync: sessions and knowledge base entries sync via Metateam's API, so agents share project memory across devices.
Data control uses .nometateam and .yesmetateam files to include or exclude directories. Session transcripts and knowledge base are stored on Metateam's API, while local artifacts (crew history, reports, config) stay on your machine.
Why It Matters
The multi-agent coding workflow is becoming a real pattern. Developers are already running Claude Code in one terminal, Gemini CLI in another, maybe Aider for a specific refactor. The friction isn't launching the agents - it's keeping track of what each one is doing and making sure they don't step on each other's work.
Metateam addresses the coordination problem. Instead of alt-tabbing between terminals and manually copying context, you get a single view of all active agents. The shared memory layer is the more interesting piece: if Agent A discovers a project convention in session 3, Agent B can access that knowledge in session 7.
For teams running AI-assisted development at scale, the multi-machine sync means a shared knowledge base across the entire team's agent sessions.
Our Take
The concept is sound - managing multiple AI coding agents is genuinely painful right now. We've run Claude Code and other tools in parallel and the context-switching overhead is real.
The persistent memory and knowledge base system is the strongest selling point here. Most agent-to-agent context sharing today is manual: you copy output from one terminal and paste it into another. Metateam tries to make that automatic.
The main concern is the API-hosted session storage. Your coding sessions contain sensitive information - file paths, code snippets, architecture decisions. SeriousBit needs to be transparent about what gets sent, how it's encrypted, and what retention policies apply. The .nometateam file helps, but the default should be minimal data transmission.
If you're already running multiple AI coding agents daily, this is worth trying. If you're using a single agent most of the time, the overhead of another tool layer probably isn't justified yet. Watch this space - the multi-agent orchestration layer is where a lot of the competition will play out in 2026.