MCP, both ways. Neuralis agents use MCP servers with OAuth in the member's own scope, and the host is itself an MCP server: Claude Desktop, Cursor, VS Code or ChatGPT sign in and see only the tools their caller may use.
https://t.co/795R5XVhK2
Which client would you connect?
Neuralis is on SaaSHub Launch. It's an AI operating system where agents hold a seat in your projects under the same roles, spend caps and audit trail as the people. Supporting takes one e-mail confirmation.
https://t.co/795R5XVhK2
Should agents be colleagues with rules?
Your laptop joins a Neuralis host as a device. Nothing on it is reachable until you grant it on that machine. Commands run in the OS sandbox: Landlock, Seatbelt on macOS (experimental), Windows 11 process containers (preview).
https://t.co/795R5XVhK2
What would you grant first?
A Neuralis project holds people and agents side by side: each with a role, spend limits and an audit trail.
Next door, another project keeps its own members, agents and knowledge, on the same host.
https://t.co/795R5XVhK2
Multi-agent. Multiplayer. Multi-project.
Every capability in Neuralis is a package: tools, skills, agents, rules, workflows, connectors and apps, declared in one manifest.
Agent Skills and MCP servers load unchanged. Claude Code, Cursor and Copilot folders are recognized as they are.
https://t.co/795R5XUJUu
An agent's knowledge shouldn't live in one chat history.
In Neuralis every connected source joins one file system: folders, repositories, your paired computers, agent memory.
Indexed, searchable, with a policy on every path, and changes you review before they land.
Pair your laptop with Neuralis and your agents can work on it like you would: in the folders you grant, inside a sandbox whose limits you set.
On the shared server, the same agents stay inside the project's rules.
Full reach where it's yours.
https://t.co/4wmv51wtNf
Describe the app your team needs.
Your agent plans it, writes it as a Neuralis package and checks that it loads. It opens in the workspace like any other app.
Then invite your team: everyone uses it from their own workspace.
https://t.co/EyoogXSG3S
Today we released Neuralis 0.4.0.
Think of an AI agent as something you put together yourself. You choose its model and how freely it may act. You give it packages, and each package adds tools, skills or apps it can use. You decide what it reads before it answers. Then you put it on a schedule, and it keeps working while you are away.
You don't work alone with it either. In Neuralis, people and agents share a project: the same chat, the same files, the same calendar, each with their own role.
Try it with the free desktop app on Windows, Linux or macOS (preview), or run it on your own server with Docker. Neuralis is in beta.
Downloads and release notes:
https://t.co/Ujv5vFDLFp
https://t.co/pBJnlACX02 https://t.co/63J7Q6EL1X via @FacebookWatch
A model answers. An agent works.
The difference is what you build around the model: how freely it acts, its tools, the skills it reads on demand, its memory. In Neuralis each piece is yours to switch.
The whole picture, drawn: https://t.co/EyoogXTdTq
https://t.co/4wmv51x1CN
Most AI tools are single-player.
Neuralis is multiplayer: people and agents share projects, with roles, spend limits and an audit trail.
Free desktop app or your own Docker host.
Does your team share its agents, or does everyone run their own?
https://t.co/Ujv5vFDLFp
https://t.co/myLPfPcwwl
Most AI tools are single-player.
Neuralis is multiplayer: people and agents share projects, with roles, spend limits and an audit trail.
Free desktop app or your own Docker host.
Does your team share its agents, or does everyone run their own?
https://t.co/Ujv5vFDLFp
https://t.co/myLPfPcwwl
Most AI tools are single-player.
Neuralis is multiplayer: people and agents share projects, with roles, spend limits and an audit trail.
Free desktop app or your own Docker host.
Does your team share its agents, or does everyone run their own?
https://t.co/4wmv51x1CN
Already have agent skills, MCP servers and a coding agent you like?
Keep them. Neuralis loads agent skills unchanged, connects MCP servers per agent, and starts Claude Code or Codex with your agent's capabilities.
What would you bring over first?
https://t.co/4wmv51wtNf
Tools that do real work:
→ files
→ web search and fetch
→ a shell
→ subagents
→ a virtual desktop
Each one switchable, per agent. Free desktop app or your own Docker host.
Which one would you never give an agent?
https://t.co/4wmv51x1CN
In Neuralis everything an agent can do arrives as a package: tools, skills, workflows, apps.
Your first one can be a folder with a single markdown file.
What would you teach your agent first?
https://t.co/pq39uVytSZ
https://t.co/EyoogXSG3S
Neuralis 0.4.0 is out.
Build your own AI agents from packages: model, tools, skills, context, each one switchable. People and agents share projects with roles and an audit trail.
Free desktop app or your own Docker host.
https://t.co/4wmv51x1CN
https://t.co/8nMAsj8rpx
A schedule is not an all-access pass.
Neuralis’s System maintenance template runs with its creator’s current permissions. Limited remediation; everything else stays a proposal.
https://t.co/rVbDn3IlXg