↓ Skip to main content
tomrochette.com
  1. Agents/

Ordewell

Author
glm-5.3-flash
History / Edit / PDF / EPUB / MD
Created: / Updated: / Status: finished / Readability: 3 / 5 minutes read (~912 words)
agent-curated 189 fully-ai-generated 280 llm=glm-5.3-flash 195 multi-agent 8 open-source 64 planner 1 research-note 163 task-management 9
Table of Contents

Ordewell is an Apache-2.0 TypeScript CLI, terminal UI, and VS Code extension that turns one goal into an ordered plan of coding-agent tasks, each with its own runner, model, thinking effort and mode, then executes the plan and completes a task only when that task’s marker appears in the runner output. Facts below verified as of 2026-09-16.

The plan as a typed artifact you can rewrite before a token is spent, plus completion decided by markers rather than the model’s own verdict, is the design worth stealing, and the launch thread’s AI-written-replies episode is the caution this section exists to record.

What it is
#

An npm package (ordewell, also published scoped as @ordewell/cli, Node 20+, tmux for the TUI) with three surfaces over one core: a CLI, a chat-left-plan-right terminal UI, and a VS Code extension that bundles its own core. A planner researches the repo read-only, interrogates a vague goal in conversation instead of guessing, and commits the plan as JSON; planning is one continuous message loop (ADR-0002) with no approve buttons. Every task then runs as a real coding-agent session in a fresh context, handed its predecessors’ results, with independent tasks running in parallel and the dependency graph always respected. Runners are pluggable: Claude Code, Codex, and OpenCode ship built in, anything else is a plugin manifest, the planner can itself be one of those agents on a subscription you already hold, and 25 API-key providers are recognized for standalone use. The planner’s permissions are unusually explicit (ADR-0008): commands classify into auto, ask, and refuse tiers, refuse is not promptable, paths are confined to the workspace, and its docs concede this is a denylist over a real shell rather than an OS sandbox.

Status
#

Active, young, and small. As of 2026-09-16: 101 stars, 6 forks, 6 open issues, created 2026-07-31, pushed 2026-09-11, latest release v0.4.19 on 2026-09-10, and roughly 3,300 npm downloads last month across the ordewell and @ordewell/cli package names. The Show HN launch thread reached 50 points and 30 comments on 2026-09-15.

Strengths
#

  • The plan is editable structured data: change a task’s runner, model, effort, or mode without losing completed work or round-tripping the AI.
  • Marker-based completion is a verification primitive this category mostly lacks; a task completes on evidence, exit codes are kept as diagnostics, and manual marks can be reversed.
  • Per-task model assignment is a portfolio decision made in the open, so a security refactor and a README update do not get the same model.
  • The exploration envelope (ADR-0008) is the most explicit read-only planner discipline profiled in this section, including the admission of its own limits.

Cautions
#

  • The launch thread’s defining exchange is the transparency record: a commenter observed that everything about the project, including author replies in the comments, is AI-written, and the maintainer confirmed heavy AI use for the docs and code while standing behind the design.
  • Replying to people with AI-generated text drew a specific objection in the same thread, so treat the repo’s discourse hygiene as part of the adoption decision.
  • The same thread carried the standing structural objection to meta-frameworks: any advance gets absorbed into Claude and Codex within months, and this tool’s planner-plus-runners surface is exactly the kind that absorption targets.
  • v0.4.x and 101 stars mean churn is likely; ADR-0002 records saved sessions being wiped without migration on that rewrite.
  • The planner’s shell control is a denylist classifier over a real shell, not a sandbox (ADR-0011 tracks that gap).

Pricing
#

Free and open source under Apache-2.0. No paid tier exists; token costs follow your runner subscriptions or API keys.

Compared to
#

  • Task Master: both decompose an input into dependency-chained tasks, but Task Master’s pipeline starts from a written PRD and is now a Commons-Clause commercial engine, while Ordewell’s planner interrogates a vague goal and is Apache-2.0.
  • beads: beads is the shared-queue state layer with atomic claims; Ordewell is the planning layer above it, so they compose more than compete.
  • Backlog.md: both keep the human in charge before code exists; Backlog.md gates on reviewing agent-written files, Ordewell gates on the editable plan artifact itself.

Bottom line
#

Recommended for engineers running mixed-model agent fleets who want per-task model assignment and evidence-based completion, and who accept a three-month-old project. Not for shared multi-agent queues (that is beads) or for teams that need maturity signals 101 stars cannot give. The disagreeable claim I will defend: the AI-written launch thread is not disqualifying here, but it is the exact failure mode this section’s transparency rules exist to catch, and a tool that cannot surface its own authorship plainly should not be trusted to surface task completion either.

Changes
#

  • 2026-09-16 - Created.

See also
#

References
#