Hey everyone, We posted a video showing how to integrate APIs into Agent Zero on the fly. Beyond just reading the docs, notice how the agent delivers the result? We instructed it to use the notify_user tool with HTML tags. This turns Agent Zero into a real-time dashboard for any tool you connect. 1. Upload Docs of any API. 2. Set Secrets globally or in your Project Secrets. 3. Enjoy your new integration. The new sleek UI is a work in progress available in the development branch of our GitHub. Watch the full workflow.

This video is quite relevant to our discussion today about the interface issues we're all facing at the moment, so I thought I'd share. Also Nate B. Jones is well worth following, if you don't already:

Hey everyone, We are at a pivotal moment. Today, we are announcing a proposal to acquire Vectal.ai in a $1.8M equity deal. Why are we proposing this? Agent Zero has the tech. Vectal has the distribution. By combining forces, we aim to transform Agent Zero from a GitHub repository into a serious, research-driven open-source company. The Proposed Leadership Change: - Jan ➡️ CTO: I want to get back to what I do best — building the system. - David (Vectal Founder) ➡️ CEO: David will handle the business, allowing the tech to scale. Governance & Voting: We are currently working on the contract and conditions.Once the proposal is finalized, we will submit it to a Community Vote. We believe this is the right path to ensure Agent Zero stays open, free, and powerful forever, but the final decision will rest with the community. Watch the full announcement video

I've been trying to get A0 to work with this MCP server for obsidian https://github.com/aaronsb/obsidian-mcp-plugin However, A0 seems to need SSE but the MCP server does not seem to provide this. Is there a way to configure A0's MCP integration to not use SSE? This is the error I see in the docker logs ERROR [mcp.client.sse] Error in sse_reader Traceback (most recent call last): File "/opt/venv-a0/lib/python3.12/site-packages/mcp/client/sse.py", line 72, in sse_reader async for sse in event_source.aiter_sse(): File "/opt/venv-a0/lib/python3.12/site-packages/httpx_sse/_api.py", line 38, in aiter_sse self._check_content_type() File "/opt/venv-a0/lib/python3.12/site-packages/httpx_sse/_api.py", line 19, in _check_content_type raise SSEError( httpx_sse._exceptions.SSEError: Expected response header Content-Type to contain 'text/event-stream', got 'application/json' curl'ing the MCP server endpoint gives me this curl -i http://<redacted>@localhost:3001/mcp HTTP/1.1 200 OK X-Powered-By: Express Access-Control-Allow-Origin: * Access-Control-Expose-Headers: Mcp-Session-Id Content-Type: application/json; charset=utf-8 Content-Length: 168 ETag: W/"a8-Q1309v6FrUz+rIK5jlNKBYa1cfc" Date: Sun, 28 Dec 2025 16:23:41 GMT Connection: keep-alive Keep-Alive: timeout=60 {"message":"MCP endpoint active","usage":"POST /mcp with MCP protocol messages","protocol":"Model Context Protocol","transport":"HTTP","sessionHeader":"Mcp-Session-Id"}dns@samsara:agent-zero$ curl -i http://<redacted>@localhost:3001/sse HTTP/1.1 404 Not Found X-Powered-By: Express

Agent Zero should be like Linux in terms of adoption. In fact, Agent Zero is mostly Linux. It is a "Smart Linux". Not only can it maintain and optimize the Linux environment, it can put it to work. I'd say that is a new kind of OS. Based on Linux old school, yes, but now something VERY different. Chalk one up for democracy. A smart OS. Excellent! But ( there's always a but ) The UX! THE UX! Its a UI only a geek's mother could love. Or developers. We need a "thin layer of cool." We need personas, at least. Some kind of Avatar would be great, but not mission critical. We need character voices though. This is where we can avoid a lot of the uncanny valley. And the UX needs to talk about building agents. It needs to take in pertinent data and synthesize a new persona almost instantly and let me TALK to it. Agents should be able to present visuals too. I suggest using a video conference scheme for the UI, where Agents can "share desktops". I'm guessing you are looking at A0 as a kind of universal framework that anyone can use to do anything, and I can understand why. I have a different approach in mind: Anyone can have any kind of agent they want... sorta. Agents that want to participate in the ecosystem/community are required meet standards. What are the constraints on the number of subagents A0 can have? From a performance standpoint, is it better to launch another A0, or it it better to add more compute to one instance? I'm trying to anticipate bottlenecks. Thanks