See what your coding agent is actually building.

Mapwright draws the structure, architecture, data model, cloud, API flows and logic of your project from the code, live, as the agent writes it. The agent reads the same map in about 600 tokens instead of exploring files — and everything it claims is checked against the code.

Open the live demo Set up in 2 minutes
$ pipx install mapwright-agent && mapwright init --connect claude-code
Mapwright — shop-api · 7 steps by the agent
Mapwright overview of an order service built by an agent

Ten views of one project, at every step

An IDE-style interface in dark or light. Click any box to inspect it, double-click to open the code, the flowchart or the call graph. Pick any earlier step and every view shows the project as it was then, with what changed outlined.

Overview

What the agent just did, layers, mechanisms, entry points, infrastructure, hotspots, plan and verification on one page.

Architecture

Modules grouped into layers — interface, services, data, integrations, UI — with imports and external packages.

Files

How the project is organised, with added and changed files marked at every step.

Classes

Classes, attributes, methods, inheritance and the classes they use.

Data model

Tables, columns, keys and relations from SQL, SQLAlchemy, Django and Prisma — even SQL inside code strings.

Cloud & infra

Containers, databases, caches, queues, buckets, functions, CI and external APIs, from Docker, Compose, Terraform, Kubernetes and SDK imports.

API & processes

Each route → handler → the functions it calls → the tables they touch.

Logic & algorithms

Every function by complexity with algorithm hints, and a flowchart of its control flow — each box links to its line.

Mechanisms

29 kinds — caching, retries, queues, auth, rate limiting, ORM, LLM calls … — with the evidence in the code.

Timeline

Every step: what was asked, what changed with diffs, how the design changed, and what did not check out.

Architecture view
Architecture: modules by layer, imports, external packages
Logic view
Logic: functions by complexity and a flowchart linked to the source lines
Cloud view
Cloud: Compose services, Terraform resources and the APIs the code calls
Timeline view
Timeline: every step with its changes and a verification warning

The agent writes code. Mapwright does the drawing.

Agents burn tokens rediscovering structure and writing summaries nobody can verify. With Mapwright the agent sends a title and an id; diagrams, icons, layouts and flowcharts come from a component library and from the code itself.

The agent works as usual

It reads the map with map instead of listing folders, keeps its plan in todo, and calls step "Add refunds" after each change. With hooks, turn ends close steps by themselves.

Mapwright reads the code

Static analysis of every file the agent touches — Python, JS/TS, SQL, Prisma, Docker, Terraform, CI — cached by content hash, versioned at every step. Nothing in your project is executed.

You both see the same map

You get interactive diagrams in the browser or your editor. The agent gets a compact text map and per-symbol answers: callers, callees, tables, history. No source code in either.

ToolThe agent sendsIt gets back
mapnothingthe whole project — layers, modules and key symbols, routes, tables, infra, mechanisms, todos, recent steps — in about 600 tokens
find · symbola namewhere it is, who calls it, what it calls, which tables it touches, when it changed
stepa short titlea receipt of what really changed: files, lines, new and changed functions, classes, tables
todo · claim · note · linktext, an id from the catalog, a symbolwhether the code backs it up

Nothing the agent says is taken at its word

What changed in a step is computed from the files. Todos need evidence that changed. Claimed mechanisms must appear in the code of the symbol named. Titles are checked against the changes. Whatever does not check out lands in Problems — and in the agent's own map.

See it in the demo
stepstep 7 “Add integration tests”: title mentions a test file, but no such file changed in this step
todo“integration tests against Postgres” marked done: app/services/orders.py has not changed since the todo was added
claimrate-limiting → post_order: no sign of rate limiting in post_order
verifiedretry-backoff → charge: backoff in the function itself

Connects to the agent you already use

One MCP server for every MCP-capable agent, hooks where the agent has them, an Agent Skill for agents that read skills, and plain file watching for everything else. mapwright connect --list shows what is installed.

Claude Code MCP + hooks + pluginCursor MCPVS Code · Copilot MCP + extensionWindsurf MCPGemini CLI MCPCodex CLI MCPopencode MCP + pluginClaude Desktop MCPCline · Roo MCP (paste)Zed MCP (paste)JetBrains AI MCP (paste)Any other agent file watching + CLI

Editors: a VS Code extension (also Cursor, Windsurf, VSCodium) opens the map beside your code and shows the current step in the status bar.

Measured, not promised

~600tokens for the whole map (configurable), no source code
< 3 KBall nine tool definitions, sent with each request
0.01 swarm start of a new process on an 84-file repo (0.8 s cold)
0runtime dependencies · everything stays on your machine

Validated with real Claude Code building a URL shortener from an empty folder over two turns: 23 todo calls, 9 steps, 8 claims and 3 map reads against 26 file edits — and a single file read. All 11 todos and all 8 claims verified against the code.

Give your agent a map.

Install, connect, open. Two minutes.