MCPMap
rogerchappel/mcpmap
Cartografe sua infraestrutura MCP local com segurança
Ver no GitHubFerramenta de inventário para configurações locais de Model Context Protocol (MCP) que mapeia servidores, comandos e dependências sem expor segredos. Útil para auditar e documentar qual infraestrutura de agents está configurada no workstation antes de delegar tarefas.
Instalação
npx convez-ai rogerchappel/mcpmap1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Auditar servidores MCP configurados localmente
- Documentar dependências de ambiente e credenciais sem expor secrets
- Detectar duplicatas, entradas obsoletas e configurações problemáticas
- Validar saúde de MCP configs antes de passar controle a agents
Veredito

MCPMap resolve um problema real: saber o que está configurado localmente sem vazar segredos. Essencial para quem trabalha com múltiplos MCP servers e precisa auditar a infraestrutura antes de dar agents mais acesso.
README
A tradução para o português deste README ainda está na fila.
mcpmap 🧭
mcpmap inventories local Model Context Protocol (MCP) config files and prints a redacted map of servers, commands, environment needs, and known tool names. It is a nosy little cartographer for your agent workstation — useful, cautious, and allergic to leaking secrets.
Why
MCP configs now live in several editor and agent app corners. Before you hand an agent more tools, it helps to know:
- which MCP servers are configured locally;
- which commands and working directories they use;
- which environment keys may contain credentials;
- where duplicate or stale entries are hiding.
mcpmap stays local. It does not upload inventories or run MCP servers unless you explicitly opt in with --allow-run.
Install
npm install -g mcpmap
# or, from a checkout:
npm install
npm run build
node dist/cli.js scan
Usage
mcpmap scan
mcpmap scan --config ./claude_desktop_config.json --format markdown
mcpmap scan --config ./mcp.json --no-defaults --format json
mcpmap doctor --config ./problematic.jsonc
mcpmap doctor --allow-run --timeout-ms 1000
Outputs
table(default): compact terminal overview.json: machine-readable inventory with redacted env values.markdown: pasteable documentation for team workstations.
Config discovery
By default, mcpmap checks common local MCP locations for Claude Desktop, VS Code, Cursor, Windsurf, and .mcp.json. Add --config / -c for explicit JSON or JSONC files. Use --no-defaults when you only want those explicit files.
Safety model
Default mode is static-only: read config files, parse them, redact secrets, and report doctor findings. --allow-run briefly starts each configured command to see if it survives a short startup window. That can execute arbitrary local commands from your config, so leave it off unless you trust the files being scanned.
Redaction covers sensitive env key names (TOKEN, PASSWORD, SECRET, API_KEY, etc.), common token-looking values, long opaque strings, and selected inline patterns. Treat output as safer, not magically declassified.
Doctor checks
mcpmap doctor currently flags:
- missing commands;
- commands not found on
PATHor as files; - relative
cwdvalues; - duplicate server names across scanned configs;
- disabled server entries;
- risky env keys or token-looking env values;
- unknown config shapes.
Examples
# Make a quick local map
mcpmap scan --format table
# Document one workstation config
mcpmap scan -c "$HOME/Library/Application Support/Claude/claude_desktop_config.json" --format markdown > MCP-MAP.md
# CI-style audit of checked-in sample config
mcpmap doctor --no-defaults -c ./examples/mcp.json --format json
The repository includes examples/mcp.json with fake values only, so you can inspect redaction and duplicate/risk output without exposing a workstation config.
Development
npm install
npm test
npm run check
npm run build
npm run smoke
bash scripts/validate.sh
Package contents
The npm package allowlist includes the runtime files plus the public support
documents needed for release review: README.md, LICENSE, SECURITY.md, CHANGELOG.md, CONTRIBUTING.md, CODE_OF_CONDUCT.md.
Run npm run package:smoke or npm pack --dry-run before publishing to
confirm those files are still present in the tarball.
Contributing
Bug reports and small, practical improvements are welcome. Please include fixture configs with secrets replaced by obviously fake values. Do not paste real tokens into issues, tests, or screenshots.
See CONTRIBUTING.md, SECURITY.md, and docs/PRD.md.
License
MIT © Roger Chappel
Verification
Run the release-readiness checks that match this package before publishing or opening a release PR.
npm run release:check- run the full release gate