@burdenoff/adaptercloud-mcp
v2026.825.1
Published
Adaptercloud MCP Server — AI assistant tools for the Adaptercloud platform via GraphQL
Maintainers
Readme
Adaptercloud MCP Server
Model Context Protocol (MCP) server for Adaptercloud — a thin shell over
@burdenoff/mcp-libs that
exposes AI-assistant tools, resources, and prompts for the Adaptercloud platform
via GraphQL.
All module logic (auth, sandbox, the adaptercloud product surface, and the shared
platform modules) lives in @burdenoff/mcp-libs. This repository only:
- Builds the MCP context with Adaptercloud product defaults.
- Selects which modules to mount.
- Selects the transport (stdio by default,
--httpopt-in) and port.
Quick Start
1. Install
For normal use:
npm install -g @burdenoff/adaptercloud-mcpFor source development:
cd ~/products/adaptercloud/adaptercloud-mcp
bun install2. Configure in Claude Code
Add to your project's .mcp.json (or ~/.claude/settings.json under mcpServers):
{
"mcpServers": {
"adaptercloud": {
"type": "stdio",
"command": "adaptercloud-mcp",
"env": {
"ADAPTERCLOUD_WORKSPACE_ENDPOINT": "http://localhost:4003/workspaces/graphql",
"ADAPTERCLOUD_GLOBAL_ENDPOINT": "http://localhost:4000/global/graphql",
"LOG_LEVEL": "error"
}
}
}
}3. Authenticate
On first use, ask Claude to call adaptercloud_auth_login — it returns a
verification URL and user code. Open the URL, enter the code, complete the login
in the browser, then ask Claude to call adaptercloud_auth_complete. Tokens are
persisted under ~/.config/adaptercloud-mcp/ (permissions 0o600) and auto-refresh
before expiry.
Mounted Modules
The shell mounts the Adaptercloud product module plus the shared platform modules
from @burdenoff/mcp-libs:
auth, adaptercloud, sandbox, rbac, organizations, tours, workspaces,
groups, tags, files, conversations, security, integrations,
channels, scheduler, export, notifications, support, products,
devportal, billing, store.
Tool / resource / prompt names are product-prefixed (adaptercloud_*).
Development
bun install # Install dependencies
bun run dev # Dev server (stdio, hot reload)
bun run server:http # HTTP transport mode
bun run test # Run wiring tests (vitest)
bun run sanity # typecheck + lint + format + build
bun run build # Production buildRun as a binary after building:
node dist/index.js # stdio transport
node dist/index.js --http --port 3030Environment Variables
Defaults point at PROD. Override per environment via ADAPTERCLOUD_* env vars:
| Variable | Default | Description |
|----------|---------|-------------|
| ADAPTERCLOUD_WORKSPACE_ENDPOINT | https://graphqlworkspaces.burdenoff.com/workspaces/graphql | Workspace gateway |
| ADAPTERCLOUD_GLOBAL_ENDPOINT | https://graphql.burdenoff.com/global/graphql | Global gateway |
| ADAPTERCLOUD_API_KEY | — | API key auth (skips device code) |
| ADAPTERCLOUD_ACCESS_TOKEN | — | Pre-set access token |
| ADAPTERCLOUD_WORKSPACE_ID | — | Pre-select workspace |
| ADAPTERCLOUD_ORGANIZATION_ID | — | Pre-select organization |
| LOG_LEVEL | info | Logging level (error/warn/info/debug) |
See @burdenoff/mcp-libs (src/core/config/store.ts) for the full list.
Publishing
Publishing is held (pre-launch). Pushing to main runs only the green gate
(lint, typecheck, test, build). Publishing @burdenoff/adaptercloud-mcp to npmjs happens
only via the manual release.yml workflow (workflow_dispatch or a v*
tag).
License
Proprietary — Copyright Burdenoff Consultancy Services Pvt. Ltd.
