Power your agent
The publisher's own command. Not a PlugMCP endpoint.
Host a copy with our keys and logs.
uvx c2pa-watermark-mcp==0.1.1Claude Code has no install link for a local command. Copy the command below and run it in a terminal.
claude mcp add io.github.csoai-org-c2pa-watermark-mcp -- uvx c2pa-watermark-mcp==0.1.1Run in a terminal.
{ "mcpServers": { "io.github.csoai-org-c2pa-watermark-mcp": { "command": "uvx", "args": [ "c2pa-watermark-mcp==0.1.1" ] } }}claude_desktop_config.json
Remote URL or a command on your machine? How the two installs differ
Run on PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
LaunchFrom the publisher
C2PA Content Credentials MCP server — embed and verify machine-readable provenance for AI-generated assets, ahead of the EU AI Act Article 50 deadline (2 December 2026).
EU AI Act Article 50 (Regulation (EU) 2024/1689) requires that AI-generated content be as AI-generated. The assertion is the C2PA spec's mandated way to do that.
C2PA Watermark & Provenance MCP server. Built with c2pa-python.
c2pa.ai_generatedThe deadline is 2 December 2026 — about 6 months out. Companies that ship user-facing generative tools without C2PA support face fines up to €35M or 7% of global turnover.
This MCP wraps the official c2pa-python
SDK (≥0.9.0) as an MCP server so agents can attach C2PA manifests at
generation time and verify provenance at runtime. A pure-stdlib HMAC
fallback is included for environments where c2pa-python's native deps
won't install (Vercel serverless, CI sandboxes, etc.).
sign_asset, verify_asset, statusc2pa.ai_generated assertionc2pa-python[evm] extrac2pa-watermark-vercel sibling repo# Core (HMAC fallback, no native deps)
pip install c2pa-watermark-mcp
# With full c2pa-python (X.509 chain)
pip install 'c2pa-watermark-mcp[c2pa]'
# For local development
git clone https://github.com/CSOAI-ORG/c2pa-watermark-mcp
cd c2pa-watermark-mcp
pip install -e .[dev]
from c2pa_watermark_mcp import sign_asset, verify_asset, status
# Server health
print(status())
# → {'server': 'c2pa-watermark-mcp', 'c2pa_python_available': False, ...}
# Sign an asset
with open("output.png", "rb") as f:
asset = f.read()
key = b"my-hmac-key-32-bytes-long-aaaaaa"
result = sign_asset(
asset_bytes=asset,
asset_mime="image/png",
claim_generator="MEOK-SDXL/1.0",
signing_key=key,
ai_generated=True,
assertions=[
{"label": "c2pa.training", "value": {"model": "sdxl-1.0", "dataset": "internal"}},
],
)
print(result["manifest_id"])
# Verify
verdict = verify_asset(asset, result["manifest"], key)
print(verdict) # {'valid': True, 'reasons': [], ...}
c2pa-watermark-mcp
# or with full c2pa-python
pip install 'c2pa-watermark-mcp[c2pa]'
c2pa-watermark-mcp
docker build -t c2pa-watermark-mcp .
docker compose up -d
# status endpoint on http://localhost:8000
See the companion repo c2pa-watermark-vercel
for a serverless wrapper exposing /sign, /verify, /status routes.
┌─────────────────────────────────────────────────────────────┐
│ c2pa-watermark-mcp │
├─────────────────────────────────────────────────────────────┤
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ sign_asset │ │verify_asset │ │ status │ │
│ │ (HMAC + │ │ (HMAC + │ │ (health + │ │
│ │ c2pa.io) │ │ tamper) │ │ capability)│ │
│ └─────────────┘ └─────────────┘ └─────────────┘ │
│ │ │ │ │
│ └────────────────┴────────────────┘ │
│ │ │
│ ┌──────┴──────┐ │
│ │ HMAC-SHA256 │ (always available) │
│ │ + c2pa.io │ (if c2pa-python installed)│
│ └──────┬──────┘ │
│ │ │
│ ┌─────────────┐ ┌──────┴──────┐ ┌─────────────┐ │
│ │ manifest │ │ signing │ │ AI-Act │ │
│ │ generator │ │ keys │ │ assertions │ │
│ └─────────────┘ └─────────────┘ └─────────────┘ │
└─────────────────────────────────────────────────────────────┘
c2pa.ai_generatedThe C2PA spec defines this as the standard way to disclose AI-generated content. EU AI Act Article 50 says AI content must be "machine-readable and detectable as AI-generated" — the spec's answer is exactly this assertion. Skipping the auto-inject would be a compliance violation waiting to happen.
Q: Does this need c2pa-python? A: No. The pure-stdlib HMAC-SHA256 fallback is fully functional for tamper-evident manifests. Install c2pa-python only if you need the full X.509 chain.
Q: Can I sign an asset that's not a PNG/JPEG?
A: Yes. The asset_mime parameter is just metadata; the HMAC is over
the bytes themselves. Works for video, audio, PDFs, etc.
Q: How do I rotate the signing key?
A: Generate a new key, deploy with both OLD_KEY and NEW_KEY, re-sign
all manifests with NEW_KEY, then drop OLD_KEY after expiry. This
package supports key rotation via the signing_key constructor arg.
Q: Is this production-ready? A: Yes for tamper-evidence (HMAC). For full X.509 C2PA chain + cryptographic non-repudiation, install c2pa-python ≥ 0.9.0.
MIT © MEOK AI Labs / CSOAI-ORG
© 2026 MCP Directory.
C2PA Watermark & Provenance MCP server. Built with c2pa-python.
{ "mcpServers": { "io.github.csoai-org-c2pa-watermark-mcp": { "command": "uvx", "args": [ "c2pa-watermark-mcp==0.1.1" ] } }}Tools
No tool list yet. Remote servers get one after the first health check.
FAQ
Copy the stdio config for Cursor, Claude Desktop, Claude Code, or VS Code on this page. It uses the publisher's install, not a PlugMCP proxy.
No required environment variable is listed on this record.
C2pa Watermark runs locally over stdio.
Last checked 2026-10-26. C2PA Watermark & Provenance MCP server. Built with c2pa-python.