Local MCP Package
@dypai-ai/mcp (v1.6.17) is the local Model Context Protocol package. It runs on your machine, reads your project files, and lets your AI agent deploy your app to production — something the hosted server can't do, because it never sees your filesystem.
You don't usually choose one or the other: most people connect to the remote MCP for editing and add the local package for deploys.
Remote vs local
| Remote (cloud) MCP | Local — @dypai-ai/mcp | |
|---|---|---|
| Where it runs | A hosted HTTP server | On your machine, via npx |
| Tools | 41 agent-facing tools | 28 tools (default profile) |
| Filesystem access | None | Reads your project files |
| Best for | In-IDE editing — tables, endpoints, auth, logs | Deploying, diffing, and syncing your code |
| Setup | Paste the JSON config | npx @dypai-ai/install |
The remote server is what the JSON MCP config connects to (see MCP Integration). The local package adds filesystem-aware deploy on top — it bundles your code and ships it.
Why two surfaces?
The cloud server can edit your backend from anywhere, but it can't read the files on your computer. The local package fills that gap: it pushes your local project to DYPAI and deploys it to production.
Install
The local package installs the same way as the rest of MCP — through @dypai-ai/install:
npx @dypai-ai/install
The installer lets you pick your client (Claude Code, Codex, Cursor, Windsurf, Antigravity, and others) and signs you in to fetch a token. Once configured, your AI agent can call the local tools directly. Run npx @dypai-ai/install --help for all options.
The 28 default tools
The local profile pairs filesystem-aware deploy tools (unique to the local package) with a subset of the shared remote tools, so your agent can both edit and ship without switching context.
Deploy & sync
These are the tools that need your filesystem — the reason the local package exists:
| Parameter | Type | Description |
|---|---|---|
dypai_push | Tool | Upload local changes to your DYPAI project |
dypai_pull | Tool | Pull the latest project state down to your local files |
dypai_diff | Tool | Show what differs between your local files and the deployed project |
dypai_validate | Tool | Validate your local workflows and config before deploying |
dypai_deploy_production | Tool | Deploy your app to production |
dypai_generate_types | Tool | Generate TypeScript types for your endpoints (endpoints.gen.ts) |
dypai_test_endpoint | Tool | Run a single endpoint locally to check it works |
bulk_upsert | Tool | Insert or update many database rows at once |
Database & storage
| Parameter | Type | Description |
|---|---|---|
execute_sql | Tool | Run SQL queries against your project database |
manage_database | Tool | Inspect and manage tables, columns, and migrations |
manage_storage | Tool | Upload, list, move, or delete files in a bucket |
generate_image_asset | Tool | Generate a managed image (billed as AI Credits) and save it to storage |
Auth, schedules & webhooks
| Parameter | Type | Description |
|---|---|---|
manage_users | Tool | Create, list, update, or delete users |
manage_roles | Tool | Create, update, or delete custom roles |
manage_schedules | Tool | Create, pause, resume, or delete scheduled runs |
manage_webhooks | Tool | Create and manage inbound webhook triggers |
Projects & versions
| Parameter | Type | Description |
|---|---|---|
list_projects | Tool | List the projects you have access to |
get_project | Tool | Get details for a single project |
create_project | Tool | Create a new project |
get_app_credentials | Tool | List configured credentials for the project |
get_endpoint_versions | Tool | View the version history of an endpoint |
manage_domain | Tool | Connect or inspect a custom domain |
Docs, search & artifacts
| Parameter | Type | Description |
|---|---|---|
list_ai_models | Tool | List the AI models available to your plan |
search_docs | Tool | Search DYPAI documentation |
search_logs | Tool | Search execution and application logs |
search_flow_templates | Tool | Find workflow templates for common patterns |
search_project_artifacts | Tool | Search project artifacts (files, blueprints, briefs) |
manage_project_artifact | Tool | Create or update a project artifact |
When to use which
Use the remote MCP for…
Editing your backend from inside your IDE — tables, endpoints, auth, logs, traces. 41 tools.
Use the local package for…
Shipping code — push/pull, diff against production, deploy, and generate types from your files. 28 tools.
In practice, your agent uses the remote tools to build and the local tools to deploy. Keep both configured and let it pick the right one for the task.