Frequently asked questions

ArchiCode is an open-source, local, graph-native software engineering harness — a desktop app for investigating, researching, planning, building, debugging, and evolving software with AI agents. The architecture graph is the shared realtime working contract between humans and agents: research conversations, implementation plans, source changes, tests, run history, and review decisions all stay connected to one durable model instead of disappearing into separate tools or ephemeral chat transcripts. You can start a project from scratch — designing the architecture before any code exists — or import an existing codebase; both are first-class flows.

Those tools are chat-first coding agents: you work in a conversation, and the session transcript is the working context. They are excellent at making changes, but the intent, boundaries, and decisions behind a change fade once the session ends. ArchiCode is graph-native instead of chat-first — architecture intent, code evidence, decisions, implementation scope, and verification state live in a durable, shared model that both you and the agents work from. Importantly, ArchiCode is a complete harness in its own right, with its own design vision — it doesn't sit on top of another tool's harness. Codex, Claude, and OpenAI-compatible or Anthropic-compatible APIs are simply interchangeable model sources; the persistent project context, orchestration, observability, and human review are all ArchiCode's own.

No. Traditional IDEs are file-first: great for editing details, but they don't give you the durable big picture needed to supervise a large, continuously changing AI-built system. ArchiCode is that big-picture workspace — you can keep using your favorite editor alongside it and configure ArchiCode to open projects in your chosen code IDE. It also ships practical project tooling for browsing files, inspecting diffs, operating Git, running local services, and debugging failures.

Yes — that's a core workflow. The importer analyzes an existing repository into evidence-backed architecture perspectives, grounded in real files, symbols, imports, calls, and source locations. From there you get functional knowledge maps, an editable 2D canvas, and multi-layer 3D exploration. When code changes outside ArchiCode, the resync pipeline reconciles the graph with deterministic, affected-scope updates and auditable reports.

Yes. Importing is not the only flow — you can create a project from nothing: design the architecture graph first (components, features, flows, dependencies, acceptance checks), research and refine it with the agent, and then let the build harness implement it. The graph-first approach means greenfield projects get the same durable source of truth, observability, and review workflow as imported ones, from the very first commit.

You can use Codex, Claude Code, OpenCode, Google Antigravity, Grok Build, or Kimi Code locally, or connect an OpenAI-compatible or Anthropic-compatible API. ArchiCode also supports project skills and MCP servers. Speech input and output can use downloadable local models, while low-latency voice conversations are available through OpenAI Realtime.

Not necessarily. You can connect your own token for any OpenAI-compatible or Anthropic-compatible API, or use your existing Codex or Claude Code subscription through the signed-in CLI installed on your machine. ArchiCode connects to whichever provider you choose and uses it for AI inference, while the architecture graph, context, orchestration, and review workflow remain in ArchiCode.

ArchiCode is local-first. Your project and its planning state live in a portable .archicode/ folder inside your repo; credentials and runtime-only data stay machine-local and out of source control. Code is only sent to the AI provider you explicitly configure — and if you run models locally, it doesn't leave your machine at all.

The graph is a shared contract for the whole team, not just between you and an agent. Portable planning state under .archicode/ is readable and Git-friendly, so architecture, flows, decisions, and ledgers travel through normal branches, reviews, and merges. New teammates onboard from the eagle view instead of spelunking through files, and agent work lands as explicit proposals, diffs, and verification evidence everyone can review.

ArchiCode can follow your system language or use English, French, Spanish, Portuguese, Simplified Chinese, or Japanese. The selected language applies to the application interface and agent output.

ArchiCode is a desktop app for macOS and Windows. Signed macOS builds and unsigned Windows builds are published on GitHub Releases. For a signed Windows installation, get ArchiCode directly from the Microsoft Store. To run from source, you need Node.js 22.12 or newer.

ArchiCode is free and open-source under the GNU Affero General Public License v3 only (AGPL-3.0-only). Commercial use, modification, distribution, and sale are permitted subject to the AGPL, including its source-availability and network-use requirements. A separate commercial license is available for organizations that need different terms, including closed-source distribution — reach out via the contact page.

ArchiCode is under active development, but its core engineering loop already works end-to-end: inspect an existing codebase, research and shape its architecture, queue scoped implementation, edit and test source, debug failures, verify outcomes, and reconcile the living graph as the code changes.

Contributions are welcome. External contributors accept an Individual Contributor License Agreement (with a corporate CLA when an employer owns the contribution) and retain ownership while granting the rights needed for ArchiCode's public AGPL and commercial editions. Start with CONTRIBUTING.md in the repository.