Hamster for Codex
The spec layer for OpenAI Codex.
Hamster writes the specs that Codex executes. Queue async tasks with structured context and get review-ready PRs back.
The problem
Codex needs more than a prompt.
Async execution demands upfront clarity
Codex runs in a sandbox without the ability to ask follow-up questions. Vague specs produce vague PRs. Every ambiguity in the input becomes a review comment or a rejection.
Feature-level work does not fit in one task
Codex works best on focused, atomic tasks. Handing it a full feature spec results in sprawling PRs that are hard to review. Someone has to decompose the work — manually, every time.
AI output drifts from team priorities
Without a clear link between product goals and Codex tasks, AI-generated PRs solve the wrong problems or implement features nobody prioritized.
Hamster structures work for Codex.
Write your intent in a Hamster brief, and AI expands it into a full spec. Generate a plan with scoped tasks. The Hamster CLI syncs everything into your repo as markdown — Codex reads the full brief, reasoning, and acceptance criteria before it starts.
Self-contained specs via the CLI
Codex runs asynchronously in a sandbox, so the specification must be complete before execution starts. The Hamster CLI syncs briefs, tasks, and plans into .hamster/ as markdown files — giving Codex the self-contained context it needs without anyone copying it across.
Atomic task decomposition
Codex excels at focused, well-bounded tasks. Hamster decomposes large features into atomic units that match the Codex execution model — one task, one PR, one clear outcome.
Brief-to-PR pipeline
Hamster dispatches tasks to Codex with structured specs attached, turning the brief-to-PR pipeline into a queue that Codex processes autonomously while you focus on higher-leverage work.
How teams use Hamster with Codex
Async feature pipeline
Before
Write a detailed prompt for Codex, submit it, wait for the PR, realize the spec missed edge cases, re-submit with corrections, wait again.
After
Write a Hamster brief. Hamster generates scoped task specs with acceptance criteria. Queue tasks to Codex. Review PRs that match the spec on the first pass.
Overnight batch execution
Before
Manually prepare 5 Codex tasks before leaving for the day. Each requires its own context-gathering and spec-writing effort.
After
Hamster breaks one brief into 5 tasks, each with structured specs. Queue them all to Codex before EOD. Review 5 PRs in the morning.
API development
Before
Prompt Codex with an endpoint description. Miss edge cases and validation rules.
After
Hamster brief captures all endpoints, schemas, and validation. Codex builds to spec with acceptance criteria baked into every task.
Related pages
Hamster CLI
Install the CLI, sync briefs and plans into your repo, and give your coding agent full project context.
Hamster for Developers
Implementation-ready task specs for developers who ship with AI agents.
Hamster for Founders
Turn product vision into shipped features with a lean team and AI execution.
Hamster for Claude Code
Give Claude Code structured context for every agentic coding session.
Turn your backlog into a Codex pipeline.
Spec it once. Let Codex execute. Review and ship.
Start shipping with Hamster