ToolExperimental

hive

Name: hive
Rating: 3.0 (443 reviews)
Author: tt-a1i

by tt-a1i

Browser-native multi-agent CLI orchestration with PTY-backed agent processes

TypeScript

Updated Jun 18, 2026

443

Stars

Forks

View on GitHub

What It Does

Runs multiple coding-focused LLM agents as real PTY processes in the browser or Node environment to collaborate on CLI workflows. Uses a lightweight team protocol to route stdin/stdout between agents (Claude Code, Codex, Gemini, OpenCode) and the user's terminal, letting each model act as an independent process. Distinctive features include local-first PTY integration and explicit process-level agentization for debugging and reproducibility. This aligns with the Emergence-Aware Monitoring Pattern Emergence-Aware Monitoring Pattern.

Avoid common pitfalls

Learn what failures to watch for

Why It Matters

As agents take on developer-facing tasks, seeing how individual models behave and fail in real CLI sessions is essential for trust. Hive exposes per-agent process boundaries and interaction logs so you can trace decisions, reproduce failures, and compare agent track records. That level of visibility makes it easier to form reputation signals and run pre-production A2A evaluation on coding workflows. Explainability

When to Use

Developers and researchers who want to prototype multi-agent coding workflows and observe per-agent behavior in terminal-based workflows. Agent Registry Pattern

Applications

Prototype multi-agent coding workflows that coordinate via a terminal interface
Capture per-agent stdin/stdout to reproduce failures and inspect agent decisions
Compare model behaviors in identical CLI tasks to build agent track records