# gray-stripe-minions-coding-agents-part1-2026-02-09

## Veille

Stripe Minions: autonomous end-to-end coding agents in production at scale

## Titre Article

Minions: Stripe's one-shot, end-to-end coding agents

## Date

2026-02-09

## URL

https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents

## Keywords

autonomous coding agents, Minions, Stripe, continuous integration, MCP tools, cloud devbox, goose, human code review, parallel execution, Ruby, Sorbet, shift-left feedback, shared rule files

## Authors

Alistair Gray

## Ton

**Profile**: Technical engineering article, descriptive and pragmatic register, high technical level

**Description**: Alistair Gray adopts the tone of an engineer sharing a mature internal experience report. The style is factual and methodical, structured around a concrete workflow rather than abstract promises. The article details the architecture with precision (PR figures, codebase size, number of MCP tools) while remaining accessible. The register is that of a corporate engineering post: sober, results-oriented, with measured pride in the team's accomplishments. Target audience: senior software engineers and engineering managers looking to deploy coding agents at scale in complex production environments.

## Pense-betes

- **Production volume**: More than 1,000 PRs generated by Minions and merged per week, with zero lines of code written by a human in these PRs
- **Complete workflow**: Slack message → autonomous agent work → PR with passing CI → human review only
- **Parallel execution**: Engineers deploy several Minions simultaneously to parallelize tasks
- **Massive codebase**: Hundreds of millions of lines of code, primarily in Ruby with type checking via Sorbet
- **Devboxes**: Isolated, pre-warmed cloud environments available in ~10 seconds — critical infrastructure for agent autonomy
- **Underlying agent**: Custom fork of goose, the open source coding agent from Block (formerly Square)
- **Centralized MCP integration**: "Toolshed" hosts more than 400 MCP tools, giving agents unified access to internal systems
- **Smart CI strategy**: Targeted test selection from a suite of more than 3 million tests, with a maximum of 2 CI cycles per Minion
- **"Shift feedback left" philosophy**: Failing lint in CI and enforcing it also in IDEs and git hooks, so agents get feedback as early as possible
- **Shared rule files**: The same rule files are used by Minions, Cursor and Claude Code — consistency between agents and human developers
- **Human-in-the-loop model**: The human doesn't write code, they review. The engineer's role shifts toward specification and validation

## RésuméDe400mots

Stripe has internally developed fully autonomous coding agents called "Minions", capable of producing complete pull requests without human intervention in writing the code. More than 1,000 PRs generated by these agents are merged every week, after human review only.

**The workflow** is remarkably simple: an engineer sends a Slack message describing the task, the Minion works autonomously in an isolated cloud environment (devbox), produces a PR with passing CI, then a human performs the review. Engineers can launch several Minions in parallel, thereby multiplying their production capacity.

**The infrastructure** rests on several pillars. Stripe's codebase represents hundreds of millions of lines, primarily in Ruby with the Sorbet type system. Devboxes, pre-warmed cloud environments available in about 10 seconds, provide each agent with an isolated, fully equipped workspace. The agent itself is a custom fork of goose, the open source coding tool created by Block.

**MCP integration** plays a central role via "Toolshed", a centralized platform hosting more than 400 MCP tools. This architecture gives agents unified access to Stripe's internal systems, letting them interact with the entire development ecosystem without specific configuration.

**The CI strategy** is particularly deliberate given a test suite of more than 3 million tests. Rather than running everything, the system intelligently selects the tests relevant to each change, with a maximum of 2 CI cycles allowed per Minion. This constraint forces the agents to produce correct code from the first attempts.

**The "shift feedback left" philosophy** is a guiding principle: lint rules that would fail CI are also enforced in IDEs and git hooks. This way, agents (like human developers) receive feedback as early as possible in the development cycle, reducing costly iterations.

**Consistency** is ensured by sharing the same rule files between Minions, Cursor and Claude Code. This approach guarantees that coding conventions are followed uniformly, regardless of the tool or agent used.

The Stripe model illustrates a mature vision of AI-assisted development: the human no longer writes code, they specify and validate. Agents become the producers of code, while engineers focus on technical direction, review and quality assurance.

## GrapheDeConnaissance

- Stripe —a_créé→ Minions (TECHNOLOGIE, 0.98)
- Minions —mesure→ plus de 1 000 PRs mergées par semaine (MESURE, 0.97)
- Minions —est_basé_sur→ goose (TECHNOLOGIE, 0.95)
- Block —a_créé→ goose (TECHNOLOGIE, 0.95)
- Stripe —utilise→ Toolshed (TECHNOLOGIE, 0.97)
- 400+ outils MCP —fait_partie_de→ Toolshed (TECHNOLOGIE, 0.96)
- Minions —utilise→ devboxes (TECHNOLOGIE, 0.95)
- devboxes —permet→ isolation et autonomie des agents (CONCEPT, 0.9)
- Stripe —utilise→ shift feedback left (METHODOLOGIE, 0.93)
- Minions —utilise→ fichiers de règles partagés (METHODOLOGIE, 0.92)
- fichiers de règles partagés —s_applique_à→ Minions (TECHNOLOGIE, 0.9)
- fichiers de règles partagés —s_applique_à→ Cursor (TECHNOLOGIE, 0.9)
- fichiers de règles partagés —s_applique_à→ Claude Code (TECHNOLOGIE, 0.9)
- Stripe —mesure→ codebase de centaines de millions de lignes (MESURE, 0.95)
- Stripe —utilise→ Sorbet (TECHNOLOGIE, 0.95)
- Minions —utilise→ maximum de 2 cycles CI (CONCEPT, 0.9)

---
Canonical: https://www.thekb.eu/en/fiches/gray-stripe-minions-coding-agents-part1-2026-02-09/