Software architecture for your agents.
Drag and drop your system architecture onto a canvas, or let your agent write Archer code to do it for you.
Works with any MCP client
- Codex
- Claude Code
- Cursor
- OpenCode
Everything your architecture needs
350+ component types
Agents, LLMs, RAG and MCP servers, AWS, GCP, Azure, Vercel, databases, queues and SaaS, each with its own glyph. Missing one? Add a custom type and it shows up in everyone's palette.
Planes and nesting
Lay components out on planes, like a frontend and a backend or a VPC. Any component can hold its own parts: double-click to open it, and breadcrumbs lead back out.
History with real diffs
Every revision is kept, human or agent. Diffs list what was added, removed or changed. Moving boxes around never counts.
Share a live link
A read-only link for anyone, and an SVG that stays current, for READMEs and docs.
Linked to your repo
Link a GitHub repo and branch. src paths open the code, and commits sit next to architecture revisions.
Teams and roles
Personal and team workspaces with owners, editors and viewers. Invite people by email.
Data inside components
Attach a schema, an API contract or notes to any component. JSON, YAML and tables are validated as you type, and your agent gets all of it.
Frequently Asked Questions
- Do I need an account to try it?
- No. You can go to the example sandboxes to try it out. But you will need to create an account to connect your agent.
- Which agents work with it?
- Any MCP client, through the hosted server: Claude Code, Cursor, Codex and more. Agents without MCP can read the language reference and write the file themselves.
- What does my agent actually get?
- The whole document, layout block included, plus a short reading guide: every component, what it does, where it lives in the code, how it connects and where it sits on your canvas.
- Can I start from an existing codebase?
- Yes. Give your coding agent the prompt at the top of this page: it connects the MCP server, then offers to map the repo into an Archer project.
- What if my agent gets it wrong?
- Agent writes go through the same validation as the canvas, so a broken document is rejected, not saved. Agent edits are badged until you review them, and every revision is kept.
- Can I use my own components?
- Yes. Add a custom type with its own name, glyph and color. It appears in the palette and validates in every project in the workspace.
Design systems, with your AI agents.
Start on the canvas, or hand your agent the prompt to connect it and map the codebase you already have.