Help center
Everything you need to connect an agent, wire up your apps, build tools, and understand what each call costs. New here? Start with Getting started.
Getting started
Register, create a workspace, connect an agent over MCP, and mint your first API key.Connections, Actions, and Smart Tools
The three building blocks you glue onto an agent: what each is, how they differ, and when to use which.Connect your agent
The MCP URL, claude CLI and mcp.json snippets, OAuth sign-in, and bundle URLs.API keys
What a tg_ key is, bearer usage against /mcp and /api, scopes, and revocation.Building tools
AI builds with create_action, the manual editor, the Playground, and rebuilds.Grounded Smart Tools
Point a prompt at your live data: a saved action runs before inference and fills {data}.Dashboards
Turn your tools into a live board: cards, refresh-all, a budgeted auto-refresh cadence, rules, and PDF export.Webhook triggers
A secret URL that runs one of your tools with every payload another service POSTs to it.Forms
Let your agent ask you a short questionnaire and wait for the answers, free.Billing and pricing
The prepaid balance, per-run pricing, build reservations, top-ups, and auto top-up.Connectors and connections
The catalog, OAuth and API key auth, AI-drafted connectors, editing and fixing them.External MCP servers
Bring npm and uvx MCP servers into your workspace, and what the sandbox allows.Security
Encryption, the secretless runner, egress guarding, the scrutinizer, tool windows, and audit.FAQ
Balance at zero, your own model key, revoking agents, run costs, training, and more.Troubleshooting
401s from /mcp, missing tools, failed builds, disconnected connections, denied runs.TypingMind + ToolGlue
One plugin for the whole workspace, per-bundle plugins, builder tools, Dynamic Context, and the MCP connector path.