The AI Development Orchestrator.
Geoffrussy is a CLI-based development platform that interviews you about your project, designs the architecture, generates executable plans, and orchestrates code generation across multiple AI models. Written in Go, distributed as a single binary, and positioned in the relaunch as one of the core delivery tools rather than the only story on the site.
# Install via Go toolchain
$ go install github.com/mojomast/geoffrussy@latest
# Or download a pre-built binary from GitHub Releases
$ geoffrussy init
✔ Config initialized at ~/.geoffrussy/config.yaml
$ geoffrussy interviewAn interactive CLI mode where Geoffrey asks probing questions to understand your project scope, constraints, and goals. No prompt engineering required — just have a conversation.
Generates complete system architecture documents with tech stack choices, data models, and component diagrams before a single line of code is written.
Breaks projects into atomic tasks organized in phases. Plans are exported as standard Markdown — portable, version-controllable, and readable by any agent.
Route planning to high-reasoning models (Claude, GPT) and execution to fast, cheap models (Llama, GLM, any Ollama-compatible model). Built-in cost tracking and rate limit monitoring.
Automated checkpoint reviews between phases to catch architectural drift, security issues, and alignment with the original design intent.
SQLite-backed state persistence with checkpoint/rollback. Summarizes project context into a context.md that lives with your repo.
Geoffrey asks clarifying questions about your project using high-reasoning models. What are we building? What's the tech stack? What are the constraints? The interview adapts based on your answers.
A comprehensive architecture document is generated — tech stack, component structure, data models, API design. This becomes the source of truth for everything that follows.
The project is broken down into phased DevPlans with 7-10 phases of 3-5 tasks each. Complex tasks get routed to stronger models; routine coding goes to fast, cheap ones.
Geoffrey orchestrates execution across configured models, tracking progress through each phase. Built-in cost tracking shows exactly what each step costs.
Multi-agent orchestration with 7 specialized workers. It proved that cheap agents could coordinate on real development tasks, even if the architecture was still gloriously chaotic.
Autonomous CLI coding agent with TUI, swarm mode, and DevPlan workflows. Bash + Python hybrid. Introduced structured planning but needed a ground-up rewrite.
Lessons from both ancestors distilled into a single Go binary. Interview mode, architecture-first planning, multi-model orchestration, and the DevUssy engine built in.
DevUssy started as a standalone Python planning pipeline and still stands on its own as one of the clearer planning systems in the catalog. Its circular development methodology has been integrated into Geoffrussy as the core planning engine.
When Geoffrussy generates a plan, it uses the DevUssy Protocol — the adaptive interview, complexity analysis, and phased plan generation that DevUssy pioneered. DevUssy provides the thinking patterns; Geoffrussy handles the execution (file I/O, terminal commands, git operations).