MCP Server
Hypertab is an AI-native programmable table built on the Model Context Protocol (MCP). 86 MCP tools let agents create tables, insert records, run smart columns, configure sources, queue approvals, and audit results.
Quick start
- 1. Create a Hypertab account, then reveal a workspace setup key in Settings → MCP connections.
- 2. Run the installer for your MCP client without putting the key in the command:
npx @hypertabai/mcp install --client claude-code --scope project- 3. Enter the setup key only at the installer's hidden terminal prompt. Do not add the key as a command flag or inline environment assignment; shell history can retain command text.
- 4. Restart the agent. It can discover the 86 registered tools; usable tools still depend on key scopes and current beta policy.
Core tool families
Tables, views, and records
hypertab_create_table, hypertab_list_views, hypertab_insert_rows, hypertab_query_rows, hypertab_upsert_rows
Smart columns and DAG runs
hypertab_add_smart_column, hypertab_run_column, hypertab_run_dag, hypertab_get_cell_states, hypertab_retry_failed_cells
Sources, webhooks, and webhook routes
hypertab_create_source, hypertab_create_webhook, hypertab_create_endpoint_from_source, hypertab_list_endpoint_logs
Project-first creation and repair
hypertab_list_projects → hypertab_create_project when needed → hypertab_create_table with project_id; hypertab_add_table_to_project is move/repair only
Agent bridge, events, approvals, and artifacts
hypertab_register_agent_client, hypertab_create_agent_run, hypertab_append_agent_event, hypertab_request_agent_approval, hypertab_attach_agent_artifact
Workspaces, settings, and utility
hypertab_list_workspaces, hypertab_get_usage_summary, hypertab_save_api_account, hypertab_import_csv, hypertab_export_csv
All tools prefixed hypertab_. Full per-tool reference with parameters and examples in our API docs.
Why MCP, not just REST
- Schema-first tool definitions, agents see Zod-validated parameter shapes upfront.
- Error messages include the fix, fuzzy-matched "did you mean 'leads'?" suggestions, not bare 404s.
- Resources for context, agents can read live table schemas without burning tokens on doc lookups.
- Standard transport, works with every MCP client without per-agent integration code.
- Activity records cover MCP calls and selected lifecycle events with the actor, action, target, and request context available to that path.
FAQ
- What is MCP (Model Context Protocol)? +
- MCP is an open protocol from Anthropic that lets AI agents call external tools through a standard interface. Hypertab exposes 86 MCP tools, your agent can create Projects and Tables, insert records, configure smart columns, and monitor pipelines without writing code.
- Which AI agents work with Hypertab MCP? +
- Clients that support Hypertab's MCP transport and tool schemas can connect. Common examples include Claude Code, Claude Desktop, Cursor, Windsurf, Cline, and VS Code Copilot Chat with MCP enabled; client behavior and setup can change by version.
- How do I connect my agent? +
- Open the Hypertab app with a workspace API key, copy the MCP config snippet from Settings, paste it into your agent's MCP config file, and restart. Your agent can then discover the 86 currently registered tools.
- How is auth handled? +
- The MCP config uses a revocable Hypertab API key. Requests pass through authenticated tenant and workspace controls, and supported key scopes restrict the allowed operations. Security and activity records cover MCP calls and selected lifecycle events, but Hypertab does not claim that every possible action or provider-side access is captured.
- Why MCP instead of a REST API? +
- REST is available too. MCP gives compatible AI clients schema-described tools and structured error responses, which can reduce custom integration code. Actual reliability depends on the client, model, prompt, scopes, and operation; Hypertab does not publish a universal error-reduction percentage.