Manifest Tutorial — Install and Configure in 5 Minutes

Get Manifest Running in 5 Minutes — A Real Walkthrough

When I first tried Manifest, I figured it'd take an afternoon to set up. Docker pull, config files, database setup... you know the drill. Turns out it's genuinely one command. Let me show you exactly how I got it running, including the little gotcha that tripped me up.

Prerequisites (What You Actually Need)

You'll need Docker installed, and that's basically it. No separate database — Manifest handles that. Here's the actual list:

• Docker (20.10+) on Linux, macOS, or Windows
• ~512MB of free RAM (1GB recommended)
• Port 2099 free on your host
• About 30 seconds of patience for the Docker pull

Step 1: One Command Installation

This is where Manifest surprised me. I expected a multi-step process, but here's the actual command from their README:

bash <(curl -sSL https://raw.githubusercontent.com/mnfst/manifest/main/docker/install.sh)

What this does: It downloads the Docker image, sets up a SQLite database in a volume, starts the web server on port 2099, and seeds the admin account prompt. About 20-30 seconds on a decent connection.

Step 2: Create Your Admin Account

Once the server is running, open http://localhost:2099. You'll see a clean login page. Click "Sign Up" and register — the very first account becomes the admin.

After logging in, you'll see the dashboard. It's surprisingly clean — no clutter, just a sidebar with Providers, Models, and Activity sections. I honestly liked how minimal it was.

If that image doesn't load (GitHub user-attachments can be flaky), here's the direct GitHub social preview:

Step 3: Add Your First Provider

This is the fun part. Go to Providers in the sidebar and click "Add Provider". You can add:

• API Key — Paste your OpenAI/Anthropic/etc. key directly
• Subscription — If you have a subscription plan with credits
• Local Model — Connect to a running Ollama instance
• Custom — Any OpenAI-compatible endpoint

I added OpenAI and Anthropic as API keys and a local Ollama model. Took about 2 minutes total.

Step 4: Route Your First Query

Now the cool part — replace your OpenAI base URL with Manifest's endpoint:

# Before (direct to OpenAI)
openai.api_base = "https://api.openai.com/v1"

# After (through Manifest)
openai.api_base = "http://localhost:2099/v1"
openai.api_key = "your-manifest-api-key"  # Get this from Settings -> API Keys

No code changes needed. Your existing OpenAI SDK code works exactly the same. Manifest intercepts the call, routes it based on your rules, and returns the response.

What I Wish I Knew Before Starting

• Port 2099 is hardcoded in the install script — if you need a different port, edit the docker-compose.yml after the first run
• SQLite is default — for production, you'll want PostgreSQL. Check their docs on swapping databases
• The API key (for your agents) is different from your login password. Generate one in Settings → API Keys
• Fallback is per-model-group — configure it carefully. I set Anthropic as fallback for OpenAI, and when OpenAI was down, it switched seamlessly

Conclusion

Manifest took me under 5 minutes from first command to routing real queries. The one-command install is refreshingly simple for such a powerful tool. If you manage multiple LLM providers or care about cost optimization, just try it — the setup cost is basically zero, and the payoff is immediate.