# Olaverse Lab

> Olaverse Lab is a freemium MCP server for ai models and llm tooling. Extract documents with per-field confidence, classify messages, detect 608 languages. It can run locally on your machine or as a hosted remote server.

- **Kind:** MCP server
- **Pricing:** Freemium
- **Platforms:** API, CLI
- **Website:** https://www.olaverselab.com
- **Repository:** https://github.com/olaverse-labs/olaverse-mcp
- **MCP Registry name:** io.github.Olaverse-Labs/olaverse-mcp
- **Categories:** AI models and LLM tooling, files and documents
- **Listing updated:** 2026-11-14

## About

Olaverse Lab builds small, task-specific models that are strongest on English and African-language text, and this server exposes its API as six tools. `extract_document` turns text, a JPEG or PNG scan, or a PDF into JSON shaped like your template and attaches a calibrated confidence to every field. Long documents are split, extracted in parallel and merged, with `get_extraction_job` collecting results still running. The README says fields scoring 0.95 or above are right about 95% of the time on the documents it was built for, and advises checking a threshold on your own sample before automating.

`decide` answers typed questions about a message, ticket or record with calibrated probabilities in 19 languages. `detect_language` picks among 608 languages, `restore_diacritics` restores tone marks and accents in 11 languages, and `generate_questions` writes search-style questions a passage answers, in 25 languages. Scanned PDFs without a text layer must be sent as page images.

Run it locally with `uvx olaverse-mcp` and an API key, or use the hosted endpoint with a bearer token. The API is usage-based with $1 of free credit; the underlying models are Apache-2.0 open weights.

## Features

- **Template-shaped extraction:** Returns JSON for your template with a confidence on each field.
- **Typed decisions:** Answers yes/no, choice or score questions with calibrated probabilities.
- **608-language detection:** Identifies language, including very short African-language text.
- **Diacritic restoration:** Restores tone marks, accents and Arabic vowel marks in 11 languages.
- **Question generation:** Creates search-style questions for a passage in 25 languages.
- **Long-document jobs:** Splits, extracts in parallel and merges, with a job id for slow runs.

## Good for

- Pull merchant, date, total and line items from a receipt photo
- Judge whether a Yoruba customer message is a refund request
- Identify the language of a short text
- Add tone marks back to unmarked Yoruba text
- Generate FAQ-style questions from a passage

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http olaverse-mcp https://mcp.olaverselab.com/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://mcp.olaverselab.com/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "olaverse-mcp": {
      "url": "https://mcp.olaverselab.com/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "olaverse-mcp": {
      "type": "http",
      "url": "https://mcp.olaverselab.com/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "olaverse-mcp": {
      "serverUrl": "https://mcp.olaverselab.com/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "olaverse-mcp": {
      "type": "http",
      "url": "https://mcp.olaverselab.com/mcp"
    }
  }
}
```

## Tools

- `extract_document`: Document to template-shaped JSON with per-field confidence
- `get_extraction_job`: Collect a long-document result
- `decide`: Typed questions with calibrated probabilities
- `detect_language`: Detect language among 608
- `restore_diacritics`: Restore tone marks and accents
- `generate_questions`: Generate search-style questions from a passage

## FAQ

### Does it need an API key?

Yes, an Olaverse API key is required; new accounts start with $1 of credit.

### How is it priced?

Usage-based, with no subscription; for example MIST-Extract-1.0 is $0.80 per million input tokens and $4.00 per million output tokens.

### Can I read scanned PDFs directly?

No, send their pages as JPEG or PNG images.

### Can I run the models myself?

The site says the production models are open-weight under Apache-2.0 on Hugging Face.

### Is Olaverse Lab free?

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

### How do I add Olaverse Lab to Claude Code?

Run this command in your terminal: claude mcp add --transport http olaverse-mcp https://mcp.olaverselab.com/mcp

### What are the best alternatives to Olaverse Lab?

The closest alternatives to Olaverse Lab are Claude Code Ultimate Guide, ASTra, CARTESIAN ECU, Invoices and SigPen.

## Alternatives

- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [ASTra](https://ai-powered.com/mcp/astra): ASTra is a freemium MCP server for files and documents.
- [CARTESIAN ECU](https://ai-powered.com/mcp/cartesian-ecu): CARTESIAN ECU is a paid MCP server for files and documents.
- [Invoices](https://ai-powered.com/mcp/invoices): Invoices is a paid MCP server for files and documents.
- [SigPen](https://ai-powered.com/mcp/sigpen): SigPen is a freemium MCP server for files and documents.
- [ConvertAudioToText](https://ai-powered.com/mcp/convertaudiototext): ConvertAudioToText is a freemium MCP server for design and media.
- [signafAIde](https://ai-powered.com/mcp/signafaide): signafAIde is a freemium MCP server for security.
- [Toolkien](https://ai-powered.com/mcp/toolkien): Toolkien is an open-source MCP server for ai models and llm tooling.

---
Source: https://ai-powered.com/mcp/olaverse-lab
