At a glance
What ships with the Apideck MCP Server today.
Configure your AI client
Add the Apideck MCP server to Claude, ChatGPT, Cursor, Windsurf, or any MCP-compatible client.
- Get your API key and app ID from the Apideck dashboard.
- Connect your apps through Vault to obtain a consumer ID.
- Add the MCP config to your AI client and restart it.
Features
Built for agents working across real SaaS data.
Normalized data models
Same schema across QuickBooks, Xero, NetSuite, BambooHR and 200+ other services — your agent learns the Apideck model once.
Managed OAuth via Vault
Apideck Vault handles authentication and token refresh automatically. No per-connector auth code in your agent.
Progressive discovery
Four meta-tools (list_tools, describe_tool, execute_tool, list_scopes) expose capabilities without bloating context — ~1,300 tokens instead of a full schema dump.
Connector targeting
Route requests to a specific downstream service via a single header. No code changes when switching connectors.
Scoped permissions
Enforce least-privilege access. Read-only agents cannot discover delete operations.
Auto-generated from OpenAPI
Built from the Apideck OpenAPI specification using Speakeasy. Always current, with proper Zod schemas.
Use cases
Where teams are putting the MCP server to work.
- Financial close automation across accounting platforms
- Employee onboarding workflows with HRIS systems
- Document processing and filing with cloud storage
- Accounts payable automation with bill matching
- Compliance audit preparation across multiple systems
- Cross-system revenue reporting and analysis
Prefer a lighter footprint?
The Apideck CLI gives AI agents shell access to the same Unified API with a ~80-token prompt — useful when MCP overhead is too heavy for your context budget.
Compare all agent integration options →