Claude Code, Codex, Cursor, Claude, ChatGPT, Gemini — whatever you already use. A coding agent runs the Norbix CLI. A chat app calls the MCP server every Norbix installation already is. Both get exactly the roles you pick, and every call lands in Logs.
A coding agent has a shell, so it runs the Norbix CLI. A chat app has none, so it calls the MCP server every Norbix installation already is. Both act as an AI service user with exactly the roles you pick.
norbix login → norbix ai init → norbix … --json --dry-run first, --yes only for destructive changes. Same commands in CI.https://hub.norbix.ai/v3/account/mcp A coding agent already runs shell commands all day. Instead of loading hundreds of tools into its context, it runs norbix like you do and reads one JSON answer per command. norbix ai init writes the agent's instructions into your repo — for Claude Code, Codex, Cursor, or all of them. Cowork and any other agent with a shell work the same way.
npm i -g @norbix.ai/cli
norbix login # browser: check the code, pick the roles, Allow
norbix ai init # --target claude | agents | cursor | allYou can even skip norbix login: an agent that finds it is not signed in shows you the link and the code, and waits for your Allow.
Anthropic's terminal agent. Signs in once, then builds your backend with norbix commands.
Set up Claude CodeOpenAI's coding agent. One flag puts the Norbix block into AGENTS.md.
Set up CodexAI-first editor. A rules file teaches its agent the Norbix CLI; MCP works too.
Set up CursorWindsurf's agent reaches Norbix through the hosted MCP server — or the CLI from its terminal.
Set up WindsurfGitHub Copilot agent mode. One mcp.json, press Start, sign in.
Set up VS CodeGoogle's terminal agent. Streamable HTTP to the hosted MCP server.
Set up Gemini CLIYour chat app has no shell, so it reaches your backend through MCP — the Model Context Protocol, the open standard AI tools use to call outside services as tools. Nothing to install or host: add the address as a connector, sign in with OAuth, pick the roles on Cloud's consent screen, press Allow.
https://hub.norbix.ai/v3/account/mcpSelf-hosted or enterprise: https://<your Hub>/v3/account/mcp. Load only some tool groups with ?toolsets=ai:database,ai:logs.
Use both: the CLI for the coding agent in your repo, MCP for the chat app on your desktop. Both act as AI service users with the roles you picked, and both are listed under Account → AI service users.
From The CLI for agents → CLI or MCP? in the docs.
Avatar → Connect your AI holds a prompt to paste into any AI tool, the Hub, API and MCP addresses, and one tab per tool with the exact commands. Managed, self-hosted or enterprise — Cloud fills in your own addresses.
Pipelines and workflow tools reach Norbix the way any program does: the SDKs, the REST API, or the same MCP server with a service-user key.
Visual workflow builder. Call Norbix from a workflow over HTTP or MCP.
Stateful agent graphs. Norbix is a tool node your graph can call.
Agent framework. Use Norbix as the backend your chains call.
Multi-agent crews. Each agent reads and writes through Norbix.
Self-hosted agent runner. Pair with Norbix for end-to-end automation.
AI workflow co-pilot. Norbix is the backend it builds against.
How a coding agent or chat app reaches Norbix, and as whom.
Copy-paste setup for every tool, a first prompt and what it returns.
norbix login, norbix ai init, and when to choose the CLI or MCP.
The address, OAuth or a key, the consent screen, roles and environments.
Every tool group per module, which ones read, change, or ask first.
The Viewer and Auditor roles, and a key for CI.
Install, sign in, every command, JSON output, profiles and sessions.
The two jobs: let an agent build your backend, and put AI into your product.
If it has a shell, it can run the Norbix CLI. If it speaks MCP, it can call the server. If it speaks HTTP, it has the API. Talk to us and we'll write the block.
Talk to us