Skip to content
MCP server

Zils

Train a small custom decision model on your examples and call it through an API

Runs locallyStatus and history
Install
Pricing
Freemium
Platforms
CLI, API, Web
Runs
Locally
Version
0.4.22
Listing updated
3 days ago

About Zils

Zils is a freemium MCP server for ai models and llm tooling. It runs locally on your machine and is started by your AI client.

Zils turns a set of checked examples into a small model for one repeatable decision, such as which team should handle a message, whether something needs review, or how well a product fits a request. It also ranks 2 to 10 options against a criterion and returns an ordered list with relative scores, which the site says are not confidence percentages.

You supply examples in a spreadsheet; Zils trains, evaluates on held-out examples, shows the mistakes and hosts accepted models behind an API. The site reports one research experiment: on 500 test conversations from the public ABCD dataset its model chose the recorded next action 79.2% of the time versus 70.6% for another decision model, and it notes this is not a guarantee for your task.

A data caution applies. On the free open tier, Zils network miners receive the training examples and are asked to delete them within 7 days, which Zils cannot verify; private runs use only Zils workers. The registry entry lists a stdio npm package, @zils/cli. Beta pricing is planned at $0.20 per million training tokens and $0.06 per million input tokens, with checkout not yet open.

Install

Run in a terminal

claude mcp add zils -- npx -y @zils/cli

Add to claude_desktop_config.json

{
  "mcpServers": {
    "zils": {
      "command": "npx",
      "args": [
        "-y",
        "@zils/cli"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "zils": {
      "command": "npx",
      "args": [
        "-y",
        "@zils/cli"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "zils": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@zils/cli"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "zils": {
      "command": "npx",
      "args": [
        "-y",
        "@zils/cli"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "zils": {
      "command": "npx",
      "args": [
        "-y",
        "@zils/cli"
      ]
    }
  }
}

What it does

Decision models
Trains a model for sorting, flagging or scoring from your examples.
Option ranking
Orders 2 to 10 options against a criterion, with optional grouped ranking.
Held-out evaluation
Shows performance on examples kept out of training, including mistakes.
Hosted API
Hosts accepted models and returns answers through an API call.
Open or private runs
Lets you choose whether network miners or only Zils workers train the model.
CLI package
Ships as the @zils/cli npm package over stdio.

Good for

  • Route support messages to the right team
  • Flag items that need human review
  • Score how well a result fits a query
  • Rank candidate actions for a customer issue

Pricing

Freemium

Open-tier training free (3 runs/day); private training $0.20 per million tokens, usage $0.06 per million input tokens; checkout not open yet

Prices change. Check zils.ai before you buy.

Zils FAQ

What data does it need?

Examples of decisions you already know how to make, paired with the right answers.

Who sees my training data?

On the open tier network miners receive it; private runs are trained only by Zils workers.

How much does it cost?

Planned beta rates are $0.20 per million training tokens and $0.06 per million input tokens; open-tier training is free, up to 3 runs a day.

Can I pay now?

Not yet; the pricing page says checkout and usage billing are not open.

Does a trained model always pass?

No. Training does not guarantee a model that meets your quality requirements.

Is Zils free?

Zils has a free plan, with paid plans for more usage or features.

How do I add Zils to Claude Code?

Run this command in your terminal: claude mcp add zils -- npx -y @zils/cli

What are the best alternatives to Zils?

The closest alternatives to Zils are Claude Code Ultimate Guide, signafAIde, Toolkien, Lugh and HypePaper.

Alternatives to Zils

  1. MCP server for searching the Claude Code Ultimate Guide from any MCP client

    Runs locally
    • Open source
    • 6.5K GitHub stars
    • AI models and LLM tooling
  2. signafAIde

    MCP server

    Local MCP server to set up, verify and check coverage of signed agent action records

    Runs locally
    • 5 tools
    • Freemium
    • Security
  3. Toolkien

    MCP server

    One MCP gateway for all your MCP servers, with code mode, permissions and audit

    Not checked yet
    • 3 tools
    • Remote
    • Open source
    • AI models and LLM tooling
  4. Lugh

    MCP server

    Write, check and dry-run Lugh agent programs with policy, budgets and approvals

    Runs locally
    • 5 tools
    • Open source
    • AI models and LLM tooling
  5. HypePaper

    MCP server

    Research papers, benchmarks, citations and topic experts for AI assistants

    • 58 tools
    • Remote
    • Freemium
    • AI models and LLM tooling