数据溯源工具包

Datafund Provenance Toolkit

Store data with cryptographic provenance on the Swarm decentralized network. Every upload is hashed, optionally signed by a notary, and can be anchored on-chain — giving you an immutable, verifiable record of what was stored, when, and by whom.

Architecture

┌─────────────────────────────────────────────────────────┐
│                    dataprovenance-app                     │
│                (Astro + React web UI)                     │
├──────────────────────────┬──────────────────────────────┤
│      SDK    │    CLI     │          MCP Server          │
│     (TS/JS) │  (Python)  │    (AI agent integration)    │
└───┬─────────┴──┬─────────┴──────────────┬───────────────┘
    │            │                        │
    ├────────────┼────────────────────────┤
    │       Chain Clients (direct)        │
    │       (viem / web3.py)              │
    ├────────────┼────────────────────────┘
    │            │                        │
    │   ┌────────┘          ┌─────────────┘
    │   │                   │
    ▼   ▼                   ▼
┌─────────────────────────────────────────────────────────┐
│                   Provenance Gateway                     │
│              (FastAPI — swarm_connect)                    │
│                                                          │
│       stamps, upload/download, notary signing            │
└──────────────────────────┬──────────────────────────────┘
                           │
                           ▼
                      Swarm Network
                  (decentralized storage)

    ── All clients also connect directly ──

              Blockchain RPC (Base Sepolia)
       DataProvenance smart contracts (Solidity)

All three clients (SDK, CLI, MCP Server) can anchor data hashes directly on-chain via their own chain client. The gateway handles Swarm storage, stamp management, and notary signing — it has no blockchain endpoints.

Components

Key Features

Via the gateway (all clients):

• Provenance metadata — every upload wraps data with SHA256 hash, timestamp, and encoding info
• Notary signing — optional EIP-191 cryptographic signature proving data authenticity
• Stamp management — purchase, extend, and monitor postage stamps using duration and size presets
• Free tier — rate-limited access for development and testing (3 requests/min)
• x402 payments — pay-per-request access using USDC on Base chain
• Integrity verification — automatic SHA256 hash check on download

Direct to blockchain (SDK, CLI, and MCP Server):

• Blockchain anchoring — register data hashes on the DataProvenance smart contract (Base Sepolia)
• Transformation lineage — record data transformations and merges, building a provenance DAG
• Storage references — bidirectional lookup between Swarm hashes and external storage identifiers

Quick Start

SDK (TypeScript)

pnpm add @datafund/swarm-provenance

import { ProvenanceClient } from '@datafund/swarm-provenance';

const client = new ProvenanceClient();
const result = await client.upload('Hello, World!', { standard: 'v1' });
console.log('Reference:', result.reference);

CLI (Python)

The CLI is not on PyPI. Install it from a release tag:

pip install "swarm-provenance-uploader @ git+https://github.com/datafund/swarm_provenance_CLI@v0.11.0"

Paid use with x402 needs the [x402] extra: pip install "swarm-provenance-uploader[x402] @ git+https://github.com/datafund/swarm_provenance_CLI@v0.11.0".

# Upload a file on the free tier
swarm-prov-upload --free upload --file /path/to/data.txt --usePool

# Download and verify
swarm-prov-upload download <swarm_hash> --output-dir ./downloads

• --free uses the gateway's free tier (rate-limited, no payment). It is a global option, so it goes before the command. Without it the gateway answers "Payment required".
• --usePool takes a ready postage stamp from the gateway's pool (a few seconds) instead of buying a new one (over a minute).

If the upload fails with "Failed to acquire stamp from pool: 429" or "No stamps available in pool", the pool can't serve you right now. The pool has a shared daily allowance that resets at 00:00 UTC. Run the same command without --usePool; it buys a new stamp, which takes a few minutes. To see whether the pool has stamps before you upload, run swarm-prov-upload --free stamps pool-status. A 429 at other steps comes from the general rate limit: wait a minute and try again.

MCP Server (AI Agents)

Option A — Docker (recommended):

docker pull ghcr.io/datafund/swarm-provenance-mcp:latest

Option B — From source:

The MCP server needs Python 3.10 or later.

git clone https://github.com/datafund/swarm_provenance_MCP.git
cd swarm_provenance_MCP
pip install -e .

Add to your Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "swarm-provenance": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "ghcr.io/datafund/swarm-provenance-mcp:latest"],
      "env": {
        "SWARM_GATEWAY_URL": "https://provenance-gateway.datafund.io",
        "CHAIN_ENABLED": "true"
      }
    }
  }
}

Or if installed from source, replace "command" / "args" with "command": "swarm-provenance-mcp".

Gateway

The public gateway is available at https://provenance-gateway.datafund.io. To self-host, see swarm_connect.

Status

This toolkit is a pre-release, meant for testing environments (not for production). Storage on Swarm is rented — data persistence depends on postage stamp validity.

Links

• datafund.io — Datafund
• github.com/datafund — all repos

License

MIT