What AI Agent Orchestration Means at Block
AI agent orchestration is the coordinated management of multiple specialised coding agents that research, plan, and modify large codebases while keeping humans in control through familiar tools and workflows. Block’s approach to automated software development centers on Goose, an internal orchestration framework, and Builderbot, a fleet of AI coding agents that sit on top of it. Instead of living inside a single IDE plugin, these coding agents run as shared infrastructure that understands hundreds of services and hundreds of millions of lines of proprietary code. Builderbot operates from within Slack and integrates with issue trackers such as Linear and Jira, so the same thread where work is discussed becomes the place where it is executed. This design lets Block wire AI-driven automation directly into day‑to‑day engineering habits rather than treating AI as a sidecar assistant.
Goose and Builderbot: Coding Agents Controlled from Slack
Goose provides the orchestration layer, while Builderbot acts as the user-facing framework that coordinates many coding agents at once. Engineers summon it by tagging @builderbot in Slack, then describe the task in natural language. From there, Builderbot performs research, plans changes, creates a branch, opens a pull request, watches CI, and iterates based on feedback from tests or reviewers. The conversation thread replaces the IDE for much of the workflow, reducing context switching between coding tools and team chat. According to Block, the system runs more than 200,000 operations a day and merges about 1,500 pull requests each week, representing roughly 15% of production code changes. Builderbot works only with source code and configuration files, not customer or payment data, aligning AI agent orchestration with strict security and compliance needs.
Monorepo Architecture as a Foundation for Automation
The move to AI-driven, enterprise development automation sits on top of a major architectural shift: Block consolidated around 450 JVM repositories into a single monorepo spanning Cash App and Square backend systems. Under the previous polyrepo model, teams managed separate services and shared libraries, which introduced dependency drift, diamond dependency problems, and stressful cross-repository rollouts. In the monorepo architecture, shared dependencies come from source, and atomic updates across many services can happen in a single commit. Block reports that the monorepo now supports about 8,800 builds per week, with p90 CI times around ten minutes on a stable main branch. This consolidated, dependency‑aware structure is well suited to coding agents, because a single orchestration system like Goose can see the full dependency graph instead of juggling disconnected repositories and inconsistent version histories.
How Automated Software Development Scales Across Services
Builderbot turns Goose’s orchestration capabilities into practical automated software development workflows. The system understands each internal service, API, and engineering convention, so an engineer on Cash App can safely trigger changes in a Square service they have never touched. Builderbot integrates with issue trackers to pick up tickets, perform cross-service code changes, and drive them through CI, all while keeping engineers in the Slack thread for oversight. Scale is the main differentiator from off‑the‑shelf coding assistants: Builderbot coordinates multiple agents to handle tasks ranging from simple bug fixes to large architectural migrations across databases and services. Block describes Builderbot as “the missing layer between AI coding tools and how engineering actually works at scale,” highlighting how coding agents in Slack can operate as a reliable orchestration surface for enterprise development automation, rather than as isolated helpers tied to a single repository.
