# devflowinc/trieve [Health: Active]

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/devflowinc/trieve/tree/main/clients/mcp-server  
**GitHub Stars:** 2722  
**npm Downloads (last month):** 67  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/devflowinc-trieve

## Description
Crawl, embed, chunk, search, and retrieve information from datasets through Trieve

## Tools
Capabilities this server exposes over MCP:

- **search** — Search through Trieve datasets using semantic, fulltext, hybrid or BM25 search with advanced filtering and highlighting options

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "trieve": {
    "command": "npx",
    "args": ["trieve-mcp-server@latest"],
    "env": {
      "TRIEVE_API_KEY": "",
      "TRIEVE_ORGANIZATION_ID": "",
      "TRIEVE_DATASET_ID": ""
    }
  }
}
```

**Requires environment variables:** `TRIEVE_API_KEY`, `TRIEVE_ORGANIZATION_ID`, `TRIEVE_DATASET_ID` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What devflowinc/trieve does

devflowinc/trieve connects MCP-compatible AI clients to the Trieve API. It gives agents a standardized way to search Trieve datasets and access dataset information associated with a Trieve organization.

The primary operation is `search`. Agents provide a query and dataset ID, then choose how matching should be performed. Supported search modes are semantic, full-text, hybrid, and BM25. Search requests can also include advanced filters, result-highlighting options, a page number, and a page size.

Datasets are exposed as MCP resources using URIs in the form `trieve://datasets/{dataset-id}`. In Claude Desktop, these resources appear as attachments corresponding to datasets available to the configured Trieve organization.

## How it works

The server runs as a command-line MCP process and communicates with the client using its configured MCP connection. When an agent needs information from a dataset, it can call the search tool, potentially issuing several searches while refining or breaking down a question.

A search request requires both a query and a dataset ID. Semantic search is the default mode. Full-text, hybrid, and BM25 modes are available when a different retrieval strategy is more suitable. Results default to page 1 with 10 results per page unless pagination values are supplied.

The server uses Trieve organization credentials to access the API. A dataset ID can be supplied for individual searches or configured as a default through the environment or command line.

## Setup and configuration

Install the package with:

```bash
npm install trieve-mcp-server
```

The executable can then be started with:

```bash
trieve-mcp-server
```

Set `TRIEVE_API_KEY` to an API key from the Trieve dashboard and `TRIEVE_ORGANIZATION_ID` to the corresponding organization ID. `TRIEVE_DATASET_ID` is optional and provides a default dataset when a request does not specify one. Command-line options can replace these environment variables:

```bash
trieve-mcp-server --api-key <your-api-key> --org-id <your-org-id> [--dataset-id <dataset-id>]
```

Command-line values take precedence over environment variables. For Claude Desktop, configure the server with an `npx` command and pass the credentials through the client configuration. Smithery also provides an installation command for Claude Desktop.

## Tools and capabilities

- Search a selected Trieve dataset.
- Select semantic, full-text, hybrid, or BM25 retrieval.
- Apply advanced filters to search requests.
- Customize result highlighting.
- Paginate results and control page size.
- Expose organization datasets as MCP resources.

## Limitations and notes

The search tool requires a dataset ID, either in the request or through the optional default dataset configuration. Access depends on a valid Trieve API key and organization ID. The documented interface is centered on searching and reading dataset information through MCP resources rather than dataset administration.

_Full upstream README: https://allmcps.com/mcp/devflowinc-trieve/readme_