ACI.dev

ACI.dev hooks more than six hundred tools into agentic IDEs and custom AI agents. Instead of writing a separate OAuth flow and API client for Google Calendar, Slack and every other service an agent touches, you hand authentication and function calling to one platform.

Access comes in two shapes, direct function calls through a lightweight Python or TypeScript SDK, or a unified Model Context Protocol server that a single configuration entry adds to an agentic IDE. Both are framework and model agnostic, so the choice of LLM and agent architecture stays open.

The platform is built around the parts of tool calling that break at scale.

• Multi-tenant authentication: built-in OAuth flows and secrets management for developers and end users alike.
• Dynamic tool discovery: agents find the right tool without flooding the context window with every option.
• Natural language permissions: capability boundaries are written in human-readable terms.
• Tool-use logging: see which tools an agent called and the trouble it ran into.
• VibeOps: point an IDE at Vercel, Supabase, Cloudflare and Sentry to handle provisioning, deployment, database configuration and debugging.

Common uses run from personal assistant chatbots to research, outbound sales and customer support agents that act on the systems they report to. Running the platform locally means starting the backend server and the frontend portal, each with its own guide. Backend, developer portal and integrations are all Apache 2.0, and a managed service exists for teams that skip the hosting.