1️⃣ Step 1: Download the Claude app

If you don't have it installed yet, download it directly from your phone's store:

• iOS: App Store
• Android: Google Play

If you are reading this on a computer, you can also go to claude.com/download and pick your phone from there.

Open the app and sign in with your Claude account (or create one if you don't have it).

2️⃣ Step 2: Open the chat and tap "+"

From the main chat screen, tap the + icon next to the text box to see the options you can add to the conversation.

Tap Connectors.

3️⃣ Step 3: Go into Connectors and tap "+" to add a new one

The first time you enter, the list of connectors is empty. Tap the + icon in the top-right corner and select Add custom connector (not "Browse connectors", which is the directory of already-published connectors).

4️⃣ Step 4: Fill in the custom connector with Apify's details

A form opens. Fill it in exactly like this (copy each value with its button and paste it into the matching field):

• Requires authentication: leave it enabled — this is what tells Claude that this server uses OAuth.
• OAuth client ID: leave it blank, it is optional and the Apify connector does not need it.

Tap Add.

5️⃣ Step 5: Connect your Apify account

Claude adds the connector to your list and will ask you to authorize it. Tap Connect: it will redirect you to the Apify account you created in step 0. Just tap Authorize and you'll return to Claude. If you hadn't registered yet, this is the moment when the authorization window will prompt you to.

✅ You're connected!

The Apify connector is enabled by default in your conversations, so there's nothing else to do. From here you have access to the LegalTech connectors — carry on with the example prompts below to try it out.

1️⃣ Step 1: Turn on Developer Mode

ChatGPT's custom connectors (like the Apify MCP) live behind an option called Developer Mode, meant for connectors that are not yet verified by OpenAI.

Go to Settings → Connectors → Advanced → Developer mode and flip the switch.

⚠️ ChatGPT flags this option as "high risk" because it lets you add connectors that are not verified by OpenAI. The Apify MCP is Apify's official connector, so you can enable it with confidence for this use.

2️⃣ Step 2: Create the Apify connector

With Developer Mode on, go to Connectors → Create and fill in the form with these values (copy each one with its button and paste it into the matching field):

• Authentication: choose OAuth — this is what tells ChatGPT that this server will ask you to sign in to Apify.
• Client ID and secret: leave them blank, the Apify connector does not need them.

Click Create.

3️⃣ Step 3: Connect with OAuth

When you click Create, ChatGPT opens the Apify authorization window. If you followed step 0 and are already signed in, you just have to click Sign in with Apify and accept the permissions — essentially, letting Apify run actors and return results inside your conversation. If you hadn't registered yet, this is the moment when the window will prompt you to.

Once authorized, the connector is linked to your ChatGPT account: you'll also have it available in the ChatGPT mobile app, without repeating the setup.

4️⃣ Step 4: Use Apify in the chat with @Apify

Open a new conversation, type @ and pick Apify from the list of connectors. From there you have access to the LegalTech connectors — carry on with the example prompts below to try it out.

If you want the long version of this guide, with more context on Developer Mode, it is in this blog post.

🧩 What we are building

You ──► AnythingLLM (chat and agent) ──► Ollama + gemma4:e4b (local)
             │
             └──► Apify MCP (cloud) ──► legaltech/cendoj ──► CENDOJ

• Ollama runs the language model on your machine.
• AnythingLLM is the chat app. It acts as the "agent" and is the one that calls the tools.
• Apify MCP is the connector that gives access to the LegalTech actors (CENDOJ, Constitutional Court, CJEU, INE, Cadastre…).

ℹ️ What leaves your computer: your conversations and any documents you share in the chat stay on your machine. The only thing that leaves is the search terms the agent sends to Apify and the public source (CENDOJ, for example); the results are public court rulings. So don't put personal or confidential data in the search itself. That is exactly the difference with Claude or ChatGPT: there, if you paste a contract or a claim, that text is sent to the model.

Hardware requirements

• RAM: 16 GB minimum; 32 GB recommended.
• Processor / GPU: Apple Silicon (M1 or later) or an NVIDIA GPU with 8 GB of VRAM at minimum; Apple M2/M3/M4 Pro or NVIDIA with 12 GB or more recommended.
• Free disk space: 15 GB minimum; 30 GB or more if you want to try other models.
• OS: macOS 13+, Windows 10/11 or Linux.

The gemma4:e4b model is about a 9.6 GB download (a bit more in memory, depending on the context window); AnythingLLM takes around 1–2 GB and Ollama around 1 GB. Without a GPU (CPU only) it works, but very slowly: it can take minutes per answer.

1️⃣ Step 1: Install Ollama and the model

This is the only step where you need the terminal. It can feel daunting if you have never opened it, but it is simple: you just copy and paste one command. If you get stuck or have any question, write to us at legaltechnologybootcamp@gmail.com.

1. Download and install Ollama from ollama.com/download.
2. Open the terminal: a window where you type commands instead of clicking buttons. You will only use it in this step.
   • Mac: press Cmd + Space, type Terminal and press Enter.
   • Windows: press the Windows key, type Terminal (or PowerShell) and press Enter.
3. Copy this command, paste it into the terminal and press Enter. It downloads the model (about 10 GB, it may take a while):

   ollama pull gemma4:e4b

4. Check that it supports tools. tools must appear in the output of this command:

   ollama show gemma4:e4b

5. Quick test (exit with /bye):

   ollama run gemma4:e4b "Hi, answer in one sentence"

🎉 Congratulations! You now have a language model running on your own computer. It doesn't depend on any company or on an internet connection: only on your machine.

2️⃣ Step 2: Install and configure AnythingLLM

Ollama runs from the terminal. AnythingLLM puts a chat window on top of it, so day to day you use it just like ChatGPT or Claude: you type, attach documents and read the answers, without opening the terminal again.

1. Download AnythingLLM Desktop from anythingllm.com/download and install it.
2. Go to Settings → AI Providers → LLM:
   • Provider: Ollama
   • Base URL: http://127.0.0.1:11434 (usually detected automatically)
   • Model: gemma4:e4b
3. Create a workspace (for example, legaltech). In its settings, under Agent Configuration, also select Ollama and gemma4:e4b as the agent model.

3️⃣ Step 3: Copy your Apify token

1. With the account from step 0 already created, go to Apify Console → Settings → API & Integrations and copy your Personal API token (it starts with apify_api_...).
2. 🔐 Treat it like a password: don't share it or push it to repositories, cover it if you take screenshots of the configuration file and, if it leaks, revoke it and create a new one from the same screen.

4️⃣ Step 4: Set up the Apify MCP connector in AnythingLLM

1. In AnythingLLM, go to Settings → Agent Skills → MCP Servers and click the book icon. A text file with the configuration opens.
2. Delete whatever it contains and paste this (use the Copy configuration button). Then replace YOUR_APIFY_TOKEN with your token from step 3, keeping the word Bearer and a space in front. Save the file.

{
  "mcpServers": {
    "apify": {
      "type": "streamable",
      "url": "https://mcp.apify.com/?tools=actors,docs,runs,storage,tasks,legaltech/cendoj,legaltech/hacienda,legaltech/tribunal-constitucional,legaltech/tjue,legaltech/fiscalia,legaltech/catastro,legaltech/ine,legaltech/aepd",
      "headers": {
        "Authorization": "Bearer YOUR_APIFY_TOKEN"
      }
    }
  }
}

Go back to Settings → Agent Skills → MCP Servers and click Refresh. "Apify" should show as On, with its list of tools.

5️⃣ Step 5: Enable only the minimum tools (important!)

The Apify connector exposes about 30 tools, and each one's description is sent to the model with every message. With all of them enabled they take up about 12,000–16,000 tokens: the model has no room left to think, so it hangs or makes up tools. On top of that, a small model gets far more confused the more options it has.

In Settings → Agent Skills → MCP Servers → Apify, leave only these three enabled:

• ✅ legaltech--cendoj: search rulings and extract their text.
• ✅ get-actor-run: check whether a long run has finished.
• ✅ get-dataset-items: fetch the results of that run.
• ❌ Everything else disabled, including call-actor: it is redundant and its description is very long.

That leaves 3/30 tools enabled. If you want another source (Constitutional Court, CJEU, INE…), enable only that legaltech--… tool and disable CENDOJ: one at a time works best. Rule of thumb: the fewer tools, the more free context and the more accurate the answers.

6️⃣ Step 6: Use it

1. Open a new chat (New Thread) in your workspace.
2. Start the first message with @agent to turn on agent mode. Later messages in the same thread stay in agent mode.
3. When the agent uses the connector you'll see something like "MCP server: apify:legaltech--cendoj completed successfully". Click the > arrow to see which parameters the model sent.

Try one of these examples (CENDOJ is a Spanish database, so the legal terms stay in Spanish):

✅ Best practices

• Open a new chat every few searches. Results take up a lot of context.
• Ask for few results (5–10). To read rulings, use paragraph mode instead of the full text.
• Always verify. Look up the ROJ on poderjudicial.es or open the PDF. The model can summarise badly or mix up data.
• If the model falls short, try a bigger one with the same setup (for example, qwen3.6:35b-a3b if you have 32 GB of RAM or more). It is slower, but it gets the parameters right more often.