ACI.dev is an open-source tool-calling platform written in Python that integrates 600+ tools into agentic IDEs and custom AI agents. The platform provides two primary access methods: direct function calling or through a unified Model-Context-Protocol (MCP) server. It addresses the complexity of managing authentication, permissions, and tool discovery at scale by offering multi-tenant authentication, granular permission controls, and dynamic tool discovery capabilities. The platform is framework and model agnostic, working with any LLM framework and agent architecture without vendor lock-in.
The core value proposition centers on improving tool-calling reliability and accountability. Rather than requiring developers to write separate OAuth flows and API clients for individual services like Google Calendar or Slack, ACI.dev centralizes authentication management and provides unified, secure function calls. The platform includes built-in OAuth flows, secrets management for both developers and end-users, and natural language permission boundaries that control agent capabilities with human-readable constraints. Tool-use logging provides visibility into how agents call tools and what issues they encounter.
The repository is actively maintained with strong community engagement.
ACI.dev emphasizes VibeOps as a key use case, enabling developers to automate DevOps tasks by providing agentic IDEs with access to platforms like Vercel, Supabase, Cloudflare, and Sentry. This allows AI to handle provisioning, deployment, database configuration, and debugging to transform prototype code into live products. Beyond DevOps automation, the platform supports personal assistant chatbots, research agents that sync findings to external apps, outbound sales agents for lead generation and CRM updates, and customer support agents that manage tickets and answer queries.
The platform is released entirely under the Apache 2.0 license, with the backend, developer portal, and integrations all open source. The ecosystem includes complementary repositories for the unified MCP server implementation, Python and TypeScript SDKs, and example agents built with ACI.dev. The project maintains comprehensive documentation, a tools directory listing available integrations, and an active community presence on Discord, Twitter, and LinkedIn. Integration requests are actively solicited through a dedicated issue template, and the project welcomes community contributions for adding new integrations or features.
