MODEL CONTEXT PROTOCOL · FREE & READ-ONLY

Pakki MCP server for packaging brand research

Ask your AI assistant which U.S. brands use valve bags, stand-up pouches, or compostable films, and get answers grounded in official sources and observation dates from the Pakki brand directories.

Create account Sign in

Updated 2026-10-20

No API key

Connect in one step. There is no account, token, or quota to manage.

Source-backed rows

Every brand result links to an official source and carries its observation date.

Evidence limits built in

Responses state what public packaging evidence can and cannot prove.

English & Chinese

Search in either language and request answers in English or Simplified Chinese.

Endpoint

The Pakki MCP server is a stateless Streamable HTTP endpoint. It answers MCP tool calls with the same data that powers the public brand directories, the JSON-LD datasets, and the CSV downloads on this site.

https://pakki.ai/api/mcp

Available tools

Available tools
Tool	What it returns	Example prompt
`search_brands`	Brands matching your keywords, with observed packaging format, official evidence, source URL, and observation date.	Find coffee roasters that use one-way valve bags.
`list_directories`	Every public directory with its entry count, last update, web page, and CSV download.	Which packaging brand directories does Pakki publish?
`get_evidence_policy`	The Observed / Derived / Inferred / Unknown labels and what public evidence cannot prove.	Can this directory tell me who supplies their pouches?

Connect your client

Pick your client and add the server. Any other client that supports remote Streamable HTTP MCP servers works with the same URL.

Claude Code

Run in your terminal:

claude mcp add --transport http pakki https://pakki.ai/api/mcp

Claude (web & desktop)

Open Settings → Connectors → Add custom connector, name it “Pakki”, and paste the URL:

https://pakki.ai/api/mcp

Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "pakki": {
      "url": "https://pakki.ai/api/mcp"
    }
  }
}

VS Code

Add to .vscode/mcp.json:

{
  "servers": {
    "pakki": {
      "type": "http",
      "url": "https://pakki.ai/api/mcp"
    }
  }
}

Evidence boundaries

The server applies the same evidence rules as the public directories.

Inclusion confirms only that an official brand source showed flexible packaging on the observation date.
Results never establish purchase intent, an open supplier search, order volume, or laminate structure.
No personal contact data is published through the website, the CSV files, or the MCP server.
Directories are curated examples, not a market sample; counts are not industry adoption rates.

Read the research methodology · Browse the brand directories

FAQ

Is the Pakki MCP server free?

Yes. The server is free, needs no account or API key, and only exposes the same public directories that are published on this website.

Does the MCP server return buyer contacts or purchase intent?

No. Results are company-level packaging observations with official sources. They never include personal contact data and do not claim that a brand is looking for a supplier.

Which MCP transport does it use?

Streamable HTTP at https://pakki.ai/api/mcp. The server is stateless and read-only, so any client that supports remote HTTP MCP servers can connect.

How is this different from the full Pakki product?

The MCP server covers the free public directories. Factory-fit scoring, dated buying signals, account discovery, and verified procurement roles are available in the Pakki app.

PACKAGING BUYER INTELLIGENCE

Go beyond public directories

Pakki connects public packaging observations with factory-fit scoring, dated buying signals, and verified procurement roles.

Create account Sign in