Releases·Tool Release
Dev builds zero-dependency Node engine for autonomous coding agents with wave execution, hash gates, and model routing
A developer posted the architecture of a zero-dependency Node.js engine that orchestrates autonomous coding agents across multiple CLIs (Claude Code, Codex, OpenCode, Gemini CLI) in local POSIX environments. It schedules tasks as parallel execution waves from markdown tables, uses content-hash snapshots to catch agents that exit cleanly without making edits, and routes tasks across model chains by role. The design targets common failure modes of agent frameworks: fake passes, unbounded token-burning loops, and heavy Python/cloud dependencies.
16 engagement1 sourceFri, Sep 25, 2026, 04:00 PM
AI Overview
- A Gate-2 content-hash check snapshots repo hashes before and after each run to catch agents that exit code 0 without making edits
- A wave-based DAG parser schedules tasks defined in markdown tables into parallel execution waves for CLIs including Claude Code, Codex, OpenCode, and Gemini CLI
- Two-tier validation uses an independent model to score task diffs (/wbValid, /wbReview) plus project-wide vertical audits (/wbAudit)
- Dynamic model routing chains per role, e.g. WORKER=qwen,gpt-5 and VALIDATOR=claude,gemini, reserving reasoning models for planning and review
- The --loop flag exits deterministically on stopping conditions: all tasks valid, a dependency block, or /wbAudit reaching a target score like >= 8/10
- Workflows replace agent personas with deterministic commands like /wbStandup, /wbSecure, /wbContext, /wbRelease, and /wbPublish
Related
Products·building
Soubi: Compiler for agent plugins across 22 coding harnesses
Soubi is a compiler that lets developers write agent plugins once in a Claude Code-compatible format and run them natively in Cursor, Codex, OpenCode, and 19 other harnesses. It solves the problem of fragmented plugin formats across AI coding tools, saving developers from manual rewrites. The tool uses LLMs to understand and translate plugin configurations.
Soubi is a compiler that lets developers write agent plugins once in a Claude Code-compatible format and run them natively in Cursor, Codex, OpenCode, and 19 other harnesses. It solves the problem of fragmented plugin formats across AI coding tools, saving developers from manual rewrites. The tool uses LLMs to understand and translate plugin configurations.
4 engagement1 sourceredditFri, Jul 31, 2026, 08:00 AM
Products·launched
Pragma: open-source terminal workspace for running coding agents across isolated Git worktrees
Pragma is a terminal-first, open-source agentic development environment for running multiple coding agents in parallel, with each task isolated in its own Git worktree. It centralizes agent status, Git/GitHub review, usage tracking, automations, and mobile access. It targets developers managing many agent sessions (Claude Code, Codex, Cursor, and others) from one place.
Pragma is a terminal-first, open-source agentic development environment for running multiple coding agents in parallel, with each task isolated in its own Git worktree. It centralizes agent status, Git/GitHub review, usage tracking, automations, and mobile access. It targets developers managing many agent sessions (Claude Code, Codex, Cursor, and others) from one place.
11120 engagement3 sourcesredditWed, Sep 23, 2026, 08:59 PM
Products·launched
Claude-B: control tmux LLM CLI sessions (Gemini CLI, Claude Code) remotely from your phone
Claude-B is a bridge that lets users send prompts to existing persistent tmux sessions running LLM coding CLIs (Gemini CLI, Claude Code, Codex, OpenCode) on their servers from a phone. Instead of moving work into a separate chat interface, it routes remote requests to the right session. It targets developers running multiple AI CLI agents on remote servers.
Claude-B is a bridge that lets users send prompts to existing persistent tmux sessions running LLM coding CLIs (Gemini CLI, Claude Code, Codex, OpenCode) on their servers from a phone. Instead of moving work into a separate chat interface, it routes remote requests to the right session. It targets developers running multiple AI CLI agents on remote servers.
4 engagement1 sourceredditThu, Sep 10, 2026, 11:11 AM
Social·Community
Developers debate whether highly autonomous AI coding workflows are practical without hundreds per month
A Reddit discussion asks whether deeply autonomous AI coding pipelines are affordable for individual developers. The poster describes a decision tree spanning harnesses (Claude Code, Codex, OpenCode, Pi/OMP), subscription tiers, and multi-model role assignment. The thread reflects practitioner uncertainty about cost-effective agentic coding setups.
A Reddit discussion asks whether deeply autonomous AI coding pipelines are affordable for individual developers. The poster describes a decision tree spanning harnesses (Claude Code, Codex, OpenCode, Pi/OMP), subscription tiers, and multi-model role assignment. The thread reflects practitioner uncertainty about cost-effective agentic coding setups.
9 engagement1 sourceredditFri, Aug 28, 2026, 08:59 AM
Products·launched
Neal: CLI for multi-agent coding with Codex and Claude
Neal is a CLI tool that orchestrates two LLM coding agents: Codex writes code and Claude reviews it. It breaks large projects into smaller chunks, starts each with fresh context to avoid context rot, and uses adversarial review to improve code quality. It solves the problem of LLM agents losing focus or degrading over long tasks.
Neal is a CLI tool that orchestrates two LLM coding agents: Codex writes code and Claude reviews it. It breaks large projects into smaller chunks, starts each with fresh context to avoid context rot, and uses adversarial review to improve code quality. It solves the problem of LLM agents losing focus or degrading over long tasks.
1 engagement1 sourcehackernewsTue, Aug 4, 2026, 01:14 PM