This MCP server lets you verify C2PA Content Credentials in media files and get LLM-ready plain-language verdicts.
Verify a local file by path or file:// URI to check if it is trusted, valid, tampered, unsigned, or has errors.
Verify a remote URL by downloading a public https media file and checking its credentials.
Scan a directory (non-recursively) to audit which files have credentials, their verdicts, signers, and AI-generated status.
Inspect server capabilities via c2pa_info: engine version, supported media types, and trust-list status.
Get structured details for each verification: signer, AI-generation indicators, edit history, provenance lineage, validation issues, ingredient issues, and trust evaluation.
Optionally include raw manifest stores to inspect the full C2PA data.
Also supports watermark declarations, trust-list status, and environment-tuned limits/security restrictions.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@c2pa-mcp verify the image at ~/Downloads/sample.jpg"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP server that verifies C2PA Content Credentials and returns an LLM-ready verdict.
Point any MCP client (Claude Desktop, Claude Code, Cursor, ...) at a local file or a URL and get back a plain-language answer: is this image/video/audio trusted, valid, tampered, or unsigned? Who signed it? Is it AI-generated? What's its edit history and provenance lineage?
Built by c2paviewer.com. Verification runs locally on the official C2PA Rust engine via @contentauth/c2pa-node. Files never leave your machine.
Read-only. This server verifies and inspects Content Credentials. It does not sign or create them.
No global install needed. Add it to your MCP client config and it runs via npx:
{
"mcpServers": {
"c2pa": {
"command": "npx",
"args": ["-y", "@c2paviewer/c2pa-mcp"]
}
}
}Claude Desktop: Settings → Developer → Edit Config, add the block above.
Claude Code: claude mcp add c2pa -- npx -y @c2paviewer/c2pa-mcp
Cursor / others: add the same mcpServers entry to the client's MCP config.
Requires Node.js 22+ (the @contentauth/c2pa-node engine requires it).
First run: the underlying engine (
@contentauth/c2pa-node) downloads a ~18MB native binary on first install. On a slow connection this can outlast an MCP client's startup timeout — if the very first launch seems to hang, let the download finish and relaunch; subsequent runs are instant.
Related MCP server: chronoverify-mcp
Tool | What it does |
| Verify a local image/video/audio/PDF by path. |
| Download a public https media URL and verify it (SSRF-guarded). |
| Audit a folder: which files have credentials, their verdict, signer, AI status. |
| Report engine version, supported media types, and trust-list status. |
Each verify tool returns a human-readable summary plus a structured digest:
{
"verdict": "valid_untrusted", // trusted | valid_untrusted | valid_trust_unknown | invalid | no_credentials | error
"summary": "Content Credentials are cryptographically valid, but the signer ... 1 earlier version in its provenance chain failed validation; this file's own signature is unaffected.",
"signer": { "name": "Example Signer", "trusted": false },
"aiGenerated": { "isAI": true, "tools": ["DALL-E"], "digitalSourceTypes": ["...trainedAlgorithmicMedia"] },
"provenance": [
{ "depth": 0, "title": "This file", "relationship": "This file", "verdict": "valid", "manifestLabel": "urn:...", "issues": [] },
{ "depth": 1, "title": "source.jpg", "relationship": "Edited from", "verdict": "invalid", "manifestLabel": "urn:...",
"issues": [ { "code": "claimSignature.mismatch", "severity": "error", "explanation": "..." } ] }
],
"edits": [ { "label": "Created", "agent": "Photoshop", "when": "...", "detail": "" } ],
"watermarks": [ { "kind": "synthid", "assertionLabel": "...", "algorithm": "" } ],
"issues": [ { "code": "signingCredential.untrusted", "severity": "warning", // THIS file's own problems (plus whatever drove an Invalid verdict)
"explanation": "The signature is cryptographically valid, but ..." } ],
"ingredientIssues": [ // problems on earlier versions in the chain
{ "ingredientTitle": "source.jpg", "manifestLabel": "urn:...", "worstSeverity": "error",
"codes": [ { "code": "claimSignature.mismatch", "severity": "error", "explanation": "..." } ] }
],
"trust": { "evaluated": true, "partial": false, "listSource": "https://.../C2PA-TRUST-LIST.pem, ..." }
}issues and verdict describe the active manifest — this file's own signature and content. Problems found on ingredients (earlier versions or components the file was made from) are reported separately in ingredientIssues and on the matching provenance[].issues, and never change the file's verdict. This follows C2PA 2.2 §15.11 (a validator must report a failing ingredient, but the active manifest keeps its own result) and matches what CAI Verify shows. So a file re-signed from a tampered original can legitimately be trusted while its lineage shows an invalid node — the model should read both.
Pass "includeRaw": true to also get the full raw manifest store.
To report a signer as trusted (not just cryptographically valid), the server checks the signing certificate against the same five trust inputs c2paviewer.com uses, fetched live and cached (24h TTL) so decisions stay current without a release:
Official C2PA Conformance Program (going-forward signers)
C2PA-TRUST-LIST.pem — CA anchors for claim signers
C2PA-TSA-TRUST-LIST.pem — CA anchors for Time-Stamp Authorities (folded into the same anchor bundle; without it valid files can carry timeStamp.untrusted)
CAI Interim Trust List (frozen Jan 2026, officially temporary, but still the only thing that recognizes pre-conformance signers — Adobe, Leica, Truepic, Canon, Samsung, Fastly, and most real-world content today)
anchors.pem — legacy CA anchors
allowed.sha256.txt — allow-list of specific end-entity certificate hashes; a signer whose leaf cert is listed is trusted even when nothing chains to an anchor
store.cfg — accepted Extended Key Usage OIDs
Without the interim group, mainstream signed content reads as valid_untrusted.
If an input can't be fetched, the server degrades loudly: with at least one anchor list loaded it still evaluates trust but sets trust.partial: true and names the missing input in trust.reason (signers recognized only by that input will read as untrusted); with no anchors at all the verdict becomes valid_trust_unknown and trust.evaluated is false. It never silently treats an unknown signer as trusted, and never silently uses a stale snapshot.
Environment overrides:
Variable | Default | Purpose |
| conformance + TSA + ITL anchors | Comma-separated PEM URLs. Replaces the default anchor set. |
| CAI | End-entity allow-list URL. Set to an empty string to disable. |
| CAI | EKU config URL. Set to an empty string to disable. |
|
| Cache lifetime for the fetched trust list. |
|
| Max download size for |
|
| Max size of a local file the file/scan tools will verify (500 MB). |
| (unset) | Comma/semicolon-separated absolute paths. When set, |
Local processing. Files are read and verified on your machine; nothing is uploaded.
SSRF-guarded URL fetching. verify_c2pa_url allows only public https, pins the resolved IP against DNS rebinding, re-validates every redirect, sends no cookies or auth, and enforces a content-type allowlist plus size cap.
Local path safety. The file/scan tools reject remote file:// and UNC paths, resolve symlinks before the allowed-roots check, and return a generic error for inaccessible paths.
Set C2PA_ALLOWED_ROOTS in any untrusted context. The file/scan tools expose the host filesystem by design, so a prompt-injected model could point them anywhere. Confining them to a specific directory is the single most important hardening setting.
Trust integrity. Trust lists are fetched over HTTPS (integrity rests on TLS and the upstream sources); the cache is per-user (0700/0600, ownership-checked) and bound to the configured URL set.
Experimental. Not legal evidence. C2PA tooling and trust infrastructure are still evolving. Do not rely on these verdicts for legal, compliance, or safety-critical decisions.
Watermarks are reported as declared, not pixel-verified. A synthid entry means the manifest declares a SynthID watermark; confirming the signal in the pixels requires the vendor's detector.
AI-generation reflects what the manifest declares via IPTC digitalSourceType. Absence of an AI declaration is not proof the content is not AI-generated.
Every verification fails with "Verification engine failed to load". The native @contentauth/c2pa-node binary was interrupted or corrupted during install (common on slow/unreliable connections). Run c2pa_info — it reports engine: FAILED: <reason> when this happens. Fix: reinstall the dependency (delete node_modules/@contentauth/c2pa-node and reinstall, or npm cache clean --force then re-run). Checksum/retry hardening of that download lives in the upstream @contentauth/c2pa-node postinstall script, which this project doesn't control.
verify_c2pa_url fails with "timed out after ...ms". The fetch exceeded C2PA_FETCH_TIMEOUT_MS (default 30000). Raise it if you're verifying URLs from a slow host.
npm install
npm run build
npm test # builds, then runs unit + end-to-end tests (network needed for the trust list)Source code is dual-licensed under MIT or Apache-2.0, at your option.
The test images under test/fixtures/ are redistributed unmodified from c2pa-org/public-testfiles and are licensed separately under CC BY-SA 4.0.
Note: if you publish under an unscoped name instead of
@c2paviewer/c2pa-mcp, changenameinpackage.jsonand theargsin the install block above; nothing else depends on the package name.
Report the verification engine version, the media types this server can verify, and the active trust-list configuration. Use this to check capabilities before verifying.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It clearly states what information is reported (version, media types, trust-list), implying a read-only, non-destructive operation. While it does not explicitly declare safety, 'Report' suggests no side effects. A score of 4 is appropriate as it is clear but could be more explicit about being read-only.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, front-loading the key information ('Report the verification engine version, the media types this server can verify, and the active trust-list configuration') followed by a usage hint. Every sentence adds value with no wasted words.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with no parameters and no output schema, the description reasonably covers what the tool does. However, it does not describe the output format or structure (e.g., JSON, fields), which would help the agent interpret results. Slight gap prevents a 5.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has zero parameters with 100% coverage, so the baseline is 4. The description adds meaning by specifying the content of the report (version, media types, trust-list), which complements the empty schema. No further parameter detail is needed.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description explicitly states the tool reports 'verification engine version, media types this server can verify, and active trust-list configuration'. The verb 'Report' clearly indicates a read operation, and the resource is well-defined. It also distinguishes from sibling tools (verify_c2pa_file, verify_c2pa_url, scan_c2pa_directory) which perform verification actions.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description includes 'Use this to check capabilities before verifying', providing explicit guidance on when to use the tool (before verification). Given the sibling tools are verification-focused, this advice is highly actionable and clarifies context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Audit a folder of media files: report which ones carry C2PA Content Credentials, their verdict, signer, and whether each is AI-generated. Non-recursive. Use this to triage a directory of images/videos at once.
| Name | Required | Description | Default |
|---|---|---|---|
| maxFiles | No | Maximum number of media files to scan (default 200, max 1000). | |
| directory | Yes | Path (or file:// URI) to the directory to scan. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries full burden. It discloses non-recursiveness and the output fields (C2PA presence, verdict, signer, AI-generation). It implies a read-only audit, but does not explicitly state no modifications or permissions needed. Slightly under-disclosed, but still useful.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences with no wasted words. First sentence defines output scope, second sentence adds usage context and a key behavioral constraint. Highly efficient.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description specifies the outputs (C2PA credentials, verdict, signer, AI-generated) without an output schema, which is adequate. Parameter count is low and well-documented in schema. Sibling tools provide context. Could mention whether it only scans media files or include a brief example, but overall complete.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% with clear descriptions for both parameters (maxFiles with default and max, directory as path or URI). The tool description adds no further parameter-level details beyond the 'non-recursive' behavioral note, which is already stated. Baseline of 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool audits a folder of media files and reports C2PA credentials, verdict, signer, and AI-generation status. It specifies non-recursive behavior, distinguishing it from sibling tools that operate on single files or URLs.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description includes an explicit usage directive ('Use this to triage a directory of images/videos at once'), providing clear context. It lacks explicit when-not-to-use guidance or alternatives, but sibling names imply differentiation.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Verify the C2PA Content Credentials (provenance) of a local image, video, audio, or PDF file. Returns a plain-language verdict (trusted / valid_untrusted / invalid / no_credentials), the signer, whether the content is AI-generated, its edit history, provenance lineage, and any validation issues. Use this when you have a file path on disk. The file is read locally and never uploaded.
| Name | Required | Description | Default |
|---|---|---|---|
| path | Yes | Absolute or relative path (or file:// URI) to the media file to verify. | |
| includeRaw | No | If true, include the full raw C2PA manifest store in the structured result. Defaults to false. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided, but the description discloses that the file is read locally and never uploaded, and enumerates the output fields (verdict, signer, AI-generated, etc.), which adds behavioral context beyond the schema.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences, front-loaded with the tool's action and outputs, followed by usage guidance. No redundancy or unnecessary details.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a tool with two parameters and no output schema or annotations, the description provides a good overview of the return values and privacy aspect. Could elaborate on the format of edit history or lineage, but overall sufficient.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the parameters are already well-documented. The description adds usage context ('Use this when you have a file path') but does not significantly augment the schema's parameter explanations.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's function: verifying C2PA content credentials of local files. It lists outputs and differentiates from verify_c2pa_url by specifying local file path usage.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Provides explicit guidance to use when there is a file path on disk, and notes the file is read locally and never uploaded. Does not explicitly mention when not to use or list alternatives, but context is clear enough.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Download a remote image, video, audio, or PDF over https and verify its C2PA Content Credentials. Returns the same plain-language verdict, signer, AI-generation status, lineage, and issues as verify_c2pa_file. Use this for a web URL. Only public https URLs are allowed (private/internal hosts are refused).
| Name | Required | Description | Default |
|---|---|---|---|
| url | Yes | Public https URL of the media file to verify. | |
| includeRaw | No | If true, include the full raw C2PA manifest store in the structured result. Defaults to false. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the burden. It discloses network access (downloading) and URL restrictions. It mentions output equivalence with verify_c2pa_file, giving insight into return format. However, it lacks details on failure modes, size limits, or performance characteristics, leaving some gaps.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is three concise sentences, each serving a purpose: action, output, usage direction, and constraint. No wasted words, and the most important info is front-loaded.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given two parameters, no output schema, and no annotations, the description covers the main behavioral points (network access, URL constraint, output similarity). It is sufficient for a tool that parallels an existing one. Minor omission: no mention of error handling or default for includeRaw (though schema covers it).
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so baseline is 3. The description adds context by linking includeRaw to the sibling's behavior, but does not elaborate beyond what the schema provides. Adequate but not enhanced.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool downloads a remote file over https and verifies C2PA credentials. It lists supported media types and explicitly distinguishes from sibling verify_c2pa_file by saying 'Use this for a web URL.' The title and description align well.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides explicit guidance: 'Use this for a web URL' and notes that only public https URLs are allowed, with private/internal hosts refused. This helps the agent decide when to use this tool. However, it does not directly mention alternatives like c2pa_info or scan_c2pa_directory, but the sibling context is available.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.3c2pa_infoscan_c2pa_directoryverify_c2pa_fileverify_c2pa_urlScored across 4 tools
Each tool has a clearly distinct purpose: local file verification, remote URL verification, directory scanning, and server info. No overlap in functionality.
Three tools use a consistent verb_c2pa_noun pattern, but c2pa_info deviates by omitting a verb. Still predictable and readable.
Four tools is on the lower end but appropriate for a focused verification server. Each tool serves a clear role without redundancy.
Covers all primary verification workflows: local file, remote URL, batch scanning, and capabilities. No obvious gaps for a verification-only tool.
Verify photos are real camera captures, not AI: C2PA and RAW+JPEG forensics. OAuth or API key.
Verify AI outputs against real sources. Returns a trust signal and permanent doc number.
Inspect and remove C2PA content credentials from AI-generated images, video and audio. Free, no key.
Verify claims with verdict, confidence & cited sources; batch verify, source checks, daily brief.
