Self-Hosted Setup Guide

Self Host Set Up

To self-host your own instance of AnotherAI, ask Claude Code to do the setup for you by sending the following prompt:

claude "Please follow instructions in https://raw.githubusercontent.com/anotherai-dev/anotherai/refs/heads/main/examples/quickstart/INSTRUCTIONS.md"

MCP Configuration

AnotherAI is available as an MCP server in the IDEs below.

Quick setup (recommended):

Complete the Self Host Set Up steps above.
Tap on this button below

Manual installation:

Open Cursor
Go to Settings...
Navigate to Cursor Settings
Select Tools and Integrations
Select + New MCP Server

Add the following configuration to your MCP servers config:

{
  "mcpServers": {
    "anotherai": {
      "url": "http://127.0.0.1:8000/mcp/"
    }
  }
}

Cursor UI should now look like this (note the green indicator and the number of tools enabled displayed):

Successful Cursor MCP Server Setup

Open Claude Code in your preferred terminal (standalone or within your IDE).
Type the following to install:

claude mcp add --scope user --transport http anotherai http://127.0.0.1:8000/mcp/

Installation Scopes: The --scope user flag installs AnotherAI for your personal use across all projects. For team projects, you can use --scope project instead to create a shared .mcp.json file that can be committed to version control and shared with your team members. Learn more about MCP scopes in the Anthropic documentation.

Type the following to verify the server is properly connected:

claude

Then type:

/mcp

You should see the AnotherAI server listed as "connected".

Open Windsurf
Go to Settings...
Select Windsurf Settings
Navigate to Cascade
Select Manage MCPs
Select View Raw Config

Add the following configuration to your MCP servers config:

{
  "mcpServers": {
    "anotherai": {
      "url": "http://127.0.0.1:8000/mcp/"
    }
  }
}

Navigate back to Manage MCPs and refresh the page. The AnotherAI MCP should appear and show as enabled.

Cursor CLI can access MCPs configured in your IDE's mcp.json configuration file, enabling the same MCP servers and tools that you've configured for the IDE.

Setup steps:

Complete the Self Host Set Up steps above and ensure the AnotherAI MCP server is running
Configure the AnotherAI MCP in your IDE's mcp.json file, as described in the Cursor tab
Once the MCP is enabled and active, you can ask the Cursor CLI to interact with it

Additional Configuration: If you find that the CLI claims it can't find or use the AnotherAI MCP, you may need to give it the specific file path to use to look for the mcp.json file.

Try it out

After you have the above set up completed, AnotherAI is ready to use!

If you have agents already created, check out how to migrate them to AnotherAI.
If you don't have any agents built yet, check out building a new agent to learn about building a new AnotherAI-compatible agent.

Once you have an agent built or migrated, here are some examples of prompts you can send to Claude Code to test how AnotherAI works:

Create an experiment in AnotherAI that compares how GPT-5 mini and GPT-4 mini perform on @[your-agent-name-here].

Use AnotherAI to help me find a faster model for @[your-agent-name-here] while still maintaining accuracy.

Configuring Authentication

The default configuration disables authentication all together using NO_AUTHORIZATION_ALLOWED=true in the .env file.

If that variable is not present, the backend expects means to validate a JWT token in environment variables, either:

JWKS_URL: the URL to a valid JSON Web Key Set
JWK: a stringified JSON of a valid JSON Web Key

You can check signature_verifier for the implementation of the signature verification and the lifecycle.py for how the verifier is selected.

The Web Client currently only supports Clerk based authentication, which is also disabled by default. It can be activated by providing the NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY and CLERK_SECRET_KEY environment variables (note that the NEXT_PUBLIC_... key must be available at build time).

Join the community

If you have questions about AnotherAI, reach out to our team and community on our community Slack.