steward.foo

THE REVIEW COMPONENT FOR SOFTWARE FACTORIES

Your software factory needs owners.

steward.foo is a fleet of narrow, always-on review agents. Each steward owns one dimension of your codebase, reviews every PR through that mission, remembers what it learns, and gets better in a loop you can inspect.

Your coding agents already produce the work. Stewards give the factory persistent review ownership: analytics, observability, security, onboarding, retention, testing—or whatever matters in your product.

Install stewardSee the loop

Production scaled. Review didn’t.

Coding agents can open PRs all day. A general reviewer can scan each diff. Neither one owns whether your analytics stay trustworthy, your telemetry keeps its contract, or your onboarding improves over time.

01

coding agents

produce changes

02

pull requests

concentrate the work

03

steward fleet

applies persistent ownership

04

merged improvements

feed the next run

memory returns to the next PR

A software factory needs more than another checkpoint. It needs persistent owners.

A different fleet for every codebase.

These are examples, not built-in checks. Each mission comes from the risks, conventions, and customer promises specific to your product.

example · PLG product

Analytics guardian

Owns event naming, funnel coverage, and telemetry drift across every release.

example · Consumer marketplace

Checkout accessibility

Owns keyboard, focus, and screen-reader regressions through purchase.

example · B2B SaaS

Customer import reliability

Owns mapping, deduplication, and recovery across every CSV import path.

Connect a repository and start with the dimensions that deserve an owner in your codebase.

Connect your repo

Every review updates the owner.

A steward does not start from zero on the next PR. It carries forward an inspectable record of what it saw, what mattered, what was accepted, and what still needs work.

  1. 01

    review

    Apply one mission to every relevant pull request.

  2. 02

    remember

    Keep findings, decisions, and unresolved work in persistent memory.

  3. 03

    improve

    Refine what matters from real outcomes and user direction.

  4. 04

    act again

    Bring accumulated context to the next pull request.

steward stateinspectable
mission
event reliability
remembered context
telemetry contracts require enforcement
open improvement
validate schema before merge
feedback trace
accepted · corrected · unresolved

You can inspect the loop. You can steer it. The owner keeps working.

See what the factory’s owners see.

This event reliability steward found a product-specific gap, then held the PR until the telemetry contract could be enforced—not merely documented.

Steward review
A steward.foo GitHub review on a pull request

The review system is part of the factory building the review system.

Install once. Review every PR.

Connect the GitHub App, define the first mission, and put a steward on the next pull request. PR review runs where the work lands; editor integrations let your coding agents define stewards and work through their improvements.

Full experience — Claude Code

Plugin, MCP auth, and the steward skills. Paste this; it does the rest.

Set up steward.foo for this repository.

Run `/plugin marketplace add contextgraph/claude-code-plugin`, then `/plugin install steward`, then `/mcp` and sign in to mcp.steward.foo when prompted.

Then run `/steward:define-steward` to define my first steward.

Any MCP client — Codex, Cursor, VS Code

Connect now and get full PR review today. In-editor commands are on the way.

{
  "mcpServers": {
    "steward": {
      "url": "https://mcp.steward.foo"
    }
  }
}

Complete the browser sign-in, install the GitHub App, and your stewards start reviewing PRs. To define a steward before in-editor commands land, ask your agent to use the steward MCP tools directly.

Give the factory its first owner.

Pick one dimension that should improve across every PR. Put a steward on it.

Install steward