For AI agents

Hibira MCP server

Let your AI coding agent set up realtime for you. Through the Model Context Protocol, Claude Code, Cursor, VS Code, Claude and ChatGPT can create apps, fill in your .env, check usage, send test events and watch the debug console.

Start freeManage AI agents

MCP server URL

https://hibira.io/mcp

  1. Create a free account

    The agent works with the apps in your Hibira account. Start free

  2. Add the server to your agent

    Run this in your project folder:

    claude mcp add --transport http hibira https://hibira.io/mcp

  3. Approve it in your browser

    The first time it connects, your agent opens a Hibira page where you sign in and choose read-only or read and write access. Agents that cannot open a browser use a personal access token instead, made on the AI agents page.

Things to ask your agent

  • “Add Hibira to this Laravel app and put the keys in .env.”
  • “Create a staging app that only accepts connections from localhost.”
  • “My chat messages are not arriving. Watch the debug console while I send one.”
  • “How close are my apps to their daily message limit?”

Tools

  • get_accountRead only

    Your plan, its account-wide limits, totals and WebSocket endpoint.

  • list_appsRead only

    Your apps with their keys, allowed origins and today’s usage. Never secrets.

  • get_appRead only

    One app’s connection details and environment variable names, with live counters on paid plans.

  • create_appRead and write

    Creates an app and returns its secret once, for your git-ignored env file.

  • update_appRead and write

    Renames an app, changes its allowed origins, or turns client events or the app on and off.

  • get_usageRead only

    Daily messages and peak connections, for the account or one app.

  • send_test_eventRead and write

    Publishes one test event. It counts toward your quota.

  • watch_debug_eventsRead only

    Watches the debug console for up to 30 seconds and returns what it saw.

  • get_integration_guideRead only

    Step-by-step instructions for your library or framework, with your app’s ID and key filled in.

Access and security

  • App secrets are shown once, when the agent creates an app. No tool reads them back.
  • Grant read-only access, or limit a personal access token to one app and an expiry date.
  • See every connected agent and token on the AI agents page, and revoke any of them at once.
  • Tool calls are limited to 5 per second per token, and test events count toward your plan like any other message.

For client developers

Remote server over Streamable HTTP (stateless, JSON responses). Sign-in is OAuth 2.1 with PKCE and dynamic client registration; metadata is at /.well-known/oauth-protected-resource. Personal access tokens are sent as a Bearer token.

All integration guides →