Comparison
Shipwright vs Factory
Factory and Shipwright agree on the shape of the problem. Both decompose a goal into a plan, put a human in front of that plan, then let agents execute and open reviewable PRs. They disagree on almost everything about how you own and operate it. Factory is a managed, enterprise-hardened, model-agnostic platform. Shipwright is an MIT harness tuned for one runtime that you run on your own infrastructure. Here is the honest, sourced comparison.
Facts verified as of August 25, 2026, from Factory's own primary sources (linked below). Re-checked immediately before this page ships: vendor terms change.
Dimension by dimension
| Dimension | Shipwright | Factory |
|---|---|---|
| License | MIT, whole harness forkable. | Commercial. |
| Deployment | Self-host: Docker or your own Kubernetes. | Cloud-managed, hybrid, or airgapped. [source] |
| Models | Claude Code, tuned end to end. | Agnostic (Claude / GPT / Gemini, BYOK). [source] |
| Plan approval | Default, on every task. | Missions flow, in their commercial offering. [source] |
| Tests-first | Enforced: tests land in the same PR, written first, CI-green to merge. | Test droid available, not an enforced gate. [source] |
| Orchestration | Shared, assignable task-store queue, cron-native. | Coordinator droid dispatches specialized droids, Mission Control view. [source] |
| Team visibility | Slack plus the PR trail. | Mission Control dashboard. |
| Enterprise security | SSO via Okta OIDC, admin vs. per-agent scoped access (cross-agent calls 403), approval gates, and an auditable review trail. | Adds SCIM auto-provisioning, compliance certifications (SOC 2, ISO), Droid Shield secret-scanning, OpenTelemetry, and airgapped deployment. [SCIM] [SOC 2 / ISO] [Droid Shield] [OpenTelemetry] [airgapped] |
| Slack workflow | First-class: the whole operating loop runs in Slack. | @-mention task delegation. [source] |
| Open source | Yes, fork and audit every line. | No. |
The architectural contrast
The resemblance is real. Factory Missions takes a business outcome in natural language, plans it, waits for you to approve the plan, then runs multiple droids that build, review, document, and test, with Git as the source of truth and a Mission Control view over the whole run. If you sketched Shipwright's loop on a whiteboard, it would look similar: plan, approve, a shared queue of work, agents that open PRs, review across the queue.
The differences are about ownership and defaults, and they are not small.
Where Factory is ahead, plainly. Shipwright already ships SSO via Okta OIDC and admin-vs-per-agent scoped access: an admin key or session unlocks everything, while a per-agent token is rejected with a 403 on any other agent's routes. Where Factory is still ahead is the rest of the enterprise trust layer: SCIM auto-provisioning, compliance certifications such as SOC 2 and ISO, Droid Shield secret-scanning before anything reaches a model, OpenTelemetry, and airgapped deployment. If you are buying for a regulated org and need those specific controls in place on day one, Factory has them today and Shipwright does not. We are not going to pretend otherwise.
Where Shipwright is ahead. Two things Factory does not do. First, tests are not optional or a droid you opt into: every task ships its tests, written before the implementation, and CI must be green before merge. Second, plan approval is the default path for all work, not a Missions mode you unlock through their commercial offering. And underneath both, Shipwright is MIT: you can fork it, read every line, and run it with nothing between your repository and the agent. Factory's plan approval and orchestration live inside a managed product you do not control.
So the choice is not "which one plans and approves." Both do. It is whether you want a managed enterprise platform with the compliance layer built, or an open harness that enforces tests-first and plan-approval by default and runs entirely on infrastructure you own.
Choose Factory when
Choose Factory when
- You are buying for an enterprise that needs SCIM auto-provisioning, compliance certifications (SOC 2, ISO), and pre-model secret-scanning in place today.
- You need airgapped or hybrid managed deployment and do not want to operate the infrastructure yourself.
- You want model flexibility across Claude, GPT, and Gemini, or BYOK.
- A multi-droid coordinator with a dedicated Mission Control dashboard is the operating model your team wants.
Choose Shipwright when
- You have committed to Claude Code and want a pipeline tuned for it, not averaged across providers.
- You want tests written first and landed with every change, gated by CI, enforced by the pipeline rather than left to an agent's judgment.
- You want plan approval on every task by default, not as a paid add-on.
- You want MIT and self-hosted, forkable and auditable, with nothing managed between your repo and the agent.
Economics
Factory's Missions are part of their commercial managed platform, sold on custom contracts. Shipwright is MIT: no seat fees and no platform charge for the harness itself. You bring your own Claude Code credentials and run it on your own Docker or Kubernetes, so your cost is your compute plus your model usage, nothing routed through an App Vitals service. The trade is the usual open-source trade: you operate it, and in exchange you own it and pay no platform margin.
Try it
Install the plugin into Claude Code and run a task end to end. Free, MIT, runs on your own infra.
/plugin install shipwright@app-vitals/shipwright