A true open source MCP platform is here! ACI.dev can directly connect to 600+ tools in one place, turning your intelligent agent into an all-rounder in seconds!

Imagine that your intelligent agent only needs to receive a natural language instruction to automatically select the most suitable API combination from 600+ tools to complete the entire task process, without the need for integration one by one or hard coding. This is no longer a future fantasy, but   a reality that has been realized by ACI.dev's open source MCP hub (Unified MCP Server).

ACI.dev is an open source infrastructure layer for AI agent tools that provides agents with intent-aware access to over 600 tools, with multi-tenant authentication, granular permissions, and dynamic tool discovery.

Using this tool, users can create an agent in just a few minutes!

Without further ado, let’s get started with the demo

In this demo, the agent completed the entire task chain with just one natural language instruction:

1️⃣ Use ACI.dev's Unified MCP Server to get the hot list of the day from Hacker News;

2️⃣ Connect to Tavily to find more context;

3️⃣ Finally, send the summary to team@aipolabs.xyz via Gmail.

There is no need for manual integration or multiple rounds of interaction throughout the process, and the tool automatically orchestrates the calls.

This is the power of Unified MCP: driven by natural language, automatically coordinated by API, and the intelligent agent directly executes the entire link with one word.

Are you interested? You can learn about or try this project through the following link:

• GitHub address: https://github.com/aipotheosis-labs/aci

• Official website address: https://www.aci.dev

This is not only a tool integration library that can be directly called, but also a unified MCP (Model Context Protocol) server :

✅ Open 600+ tool integrations at one time

✅ Dynamic discovery capabilities

✅ Multi-tenant authentication support

✅ Natural language permission control

Some of the tools and platforms supported by ACI.dev

Whether it is manually calling functions or accessing based on the MCP protocol, ACI.dev has opened the door to the world of intelligent agents for developers.

From "intention" to "action", only one step is needed

The core capability of ACI.dev is to allow the LLM agent to directly close the loop from ?️ "clarifying task objectives" → ⚙️ "accurately selecting tools to execute" , truly achieving intent-driven actions.

Unlike many projects on the market that are full of JSON schema, ACI.dev only exposes two super common meta functions:

• ACI_SEARCH_FUNCTIONS : Ask questions in natural language, and the platform will tell you which function to use;

• ACI_EXECUTE_FUNCTION : Direct execution, such as searching for papers, sending Slack messages, creating Zoom meetings, etc.

With the official Python SDK, calling these functions is as easy as drinking water.

After connecting to the MCP server, the standardized tools are ready to use without having to predefine hundreds of interfaces yourself.

And behind the scenes, ACI.dev intelligently sorts by semantic similarity to return only the most relevant tools:

• Provincial Token

• Accuracy

• Ensure complete functionality

Community developer testing

Get started with scientific research in five minutes to help intelligent agents: ArXiv × GitHub × Gmail in one go

Find some papers on reinforcement learning, send me the links to the relevant GitHub projects, and then arrange a Zoom meeting to discuss them.

Agents connected to ACI.dev can automatically complete:

• Search Arxiv for the latest papers

• Find relevant repositories on GitHub

• Send summary email or schedule Zoom meeting

The entire process is completed in one round of reasoning, without cumbersome OAuth processes or interface glue projects , making LLM truly a one-stop process.

In this presentation, we specifically cited a demo created by  Santiago Valdarrama (well-known AI engineer, X platform @svpino). Santiago is very active in the field of intelligent agents and tool calls, and is good at turning cutting-edge technologies into real cases that can be experienced in a simple and intuitive way. The following demo is a practical demonstration of intelligent agent multi-tool calls created by him based on ACI.dev Unified MCP Server.

Why open source? Why now?

In the future, SaaS vendors will also launch their own MCP Servers, and standardization of tool calls is only a matter of time.

The faster an ecosystem runs, the more it needs a layer of completely open and visible infrastructure.

ACI.dev chose to be fully open source (Apache 2.0), allowing developers to:

✅ Free Fork, customized deployment

✅ Security audit logic

✅ Self-built firewall service, no authorization required

✅ Replace underlying indexes or modules as needed

Not only for "transparency", but also for:

• Accelerate the popularization of MCP standards - ACI.dev is one of the MCP reference implementations

• Filling the existing protocol gaps - multi-tenancy, permission granularity, dynamic discovery, ACI.dev has polished it

•  Accelerate the ecosystem together - the popularization of intelligent bodies cannot be achieved by fighting alone, but by open source collaboration