Explicit plans
Structured task graphs keep agents aligned and auditable.

Open GSD keeps agentic work from dissolving into context bloat: explicit plans, clean execution contexts, real verification, and git history that tells the truth.
Built for agentic engineering
GSD is not another chat prompt. It is an operating loop for getting useful work out of AI coding tools without losing scope, state, or verification.
Structured task graphs keep agents aligned and auditable.
Right-sized context for every step. No bloat, no drift.
Automated checks and human-readable evidence you can trust.
Use the framework where you already work, the CLI when you want orchestration, the desktop app when you want visible state, and the cloud when project state needs to follow you.
Drop the GSD loop into Claude Code, Codex, Gemini CLI, Cursor, Windsurf, Copilot, or another existing runtime.
A terminal-native harness for autonomous milestones, context hygiene, worktrees, checks, cost, and recovery.
A local desktop workspace for projects, threads, plans, backlog, evidence, and ship handoffs.
$Desktop app, download coming soonComing soon
A managed workspace for cross-device project state, remote milestones, and team visibility.
$Hosted cloud, coming soonComing soon
gsd-browser is the companion product for deterministic Chrome control: MCP tools, semantic actions, assertions, recordings, traces, visual evidence, and human takeover when the work calls for it.
verification backend@opengsd/gsd-browser
$ gsd-browser mcp
navigate → snapshot → act → assert → export
01
Heavy research, planning, and implementation work happen in scoped contexts instead of one exhausted session.
02
Plans, current position, decisions, and handoffs become durable project artifacts the next run can load.
03
The work is checked at the behavior level, evidence is captured, and failures turn into actionable repair plans.