Introduction
Dexby lets your AI agent use your users' apps, with their own accounts.
Your agent asks Dexby for tools. Your user connects their Slack, GitHub or Stripe account once. Dexby stores the credential encrypted and runs tool calls with that user's connected accounts. Your project's request log settings decide what is recorded.
const tools = await dexby
.createSession({ userId: 'user_123', connectors: ['slack'] })
.andThen((session) => session.tools())Pick how your agent connects
| You use | Start with |
|---|---|
| An agent framework (AI SDK, Mastra, LangChain, OpenAI Agents and others) | Frameworks |
| A model API directly (OpenAI, Anthropic, Gemini, Bedrock) | Frameworks |
| An MCP client (Claude Code, Cursor, Claude Desktop) | MCP |
| Your own code, one tool at a time | Run a tool |
| Any language, over HTTP | REST API |
What you get
- 1,500+ connectors, each with ready tools. Browse them on the connectors page.
- Connect links: a hosted page where your users connect their own accounts. See Connect accounts.
- Sessions: one call scopes what an agent may see and do for one user. The model gets a few search and run tools instead of thousands of definitions. See Sessions.
- Triggers and webhooks: app events reach your server. See Triggers.
- Audit and erasure for every credential access and call. See Security.
- Dexby Cloud today, your own servers soon. See Self-hosting.