Announcing Beacon: Self-improving memory for coding agents

← All MCP servers

MCP server

io.github.vicseeai/vicsee-mcp-server

mcp:vicseeai/vicsee-mcp-server

Generate, edit and upscale AI video and images from any agent via VicSee.

awesome-mcp-servers Glama MCP Registry

Verdict

No issues found

No suspicious or malicious behavior was found.

Analyzed Oct 26, 2026, 12:22 UTC at e070654

0

findings

—

highest severity

0

threat classes

17

files scanned

325/325

checks satisfied

0

package advisories

Why this verdict

benign: 325 checks satisfied, no findings for mcp-server:.

Security findings 0

No security findings

No suspicious or malicious behavior was found; 3 candidate findings were refuted on review.

Execution chain 31

What it calls, runs, installs, reads and writes, with the line that does it.

Environment variables 2

VICSEE_API_KEY

src/index.ts:92 · read

VICSEE_BASE_URL

src/index.ts:93 · read

Files read 2

<dynamic>

src/index.ts:52 · read

<dynamic>

src/index.ts:131 · read

GitHub repositories 4

vicseeai/vicsee-mcp-server

package.json:32 · url

vicseeai/vicsee-mcp-server

package.json:35 · url

vicseeai/vicsee-mcp-server

package.json:38 · url

vicseeai/vicsee-mcp-server

server.json:7 · url

Network calls 6

<dynamic>

src/index.ts:149 · call

<dynamic>

src/vicsee-client.ts:82 · call

<dynamic>

src/worker.ts:64 · call

<dynamic>

src/worker.ts:160 · via:resolveUserKey

<dynamic>

src/worker.ts:207 · call

vicsee.com

src/vicsee-client.ts:8 · url_literal

Packages 1

npm:@vicsee/mcp-server@0.3.3

server.json:14 · launch:server.json

URLs 16

https://github.com/vicseeai/vicsee-mcp-server

https://github.com/vicseeai/vicsee-mcp-server.git

package.json:35

https://github.com/vicseeai/vicsee-mcp-server/issues

package.json:38

https://github.com/vicseeai/vicsee-mcp-server#readme

package.json:32

https://mcp.vicsee.com

https://mcp.vicsee.com

src/worker.ts:181

https://mcp.vicsee.com/mcp

src/worker.ts:181

https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json

https://vicsee.com

https://vicsee.com

src/vicsee-client.ts:62

https://vicsee.com/api/v1

https://vicsee.com/api/v1

src/vicsee-client.ts:2

https://vicsee.com/api/v1

src/vicsee-client.ts:8

https://vicsee.com/api/v1/internal/mcp/resolve-user-key

src/worker.ts:42

https://vicsee.com/connect/authorize

src/worker.ts:40

https://vicsee.com/mcp

Refuted findings 3

Candidates the review stage examined and rejected, kept for transparency.

Data exfiltration

criticalRefutedlow confidence

The vicsee_upload tool reads an agent-supplied local file path and uploads its bytes to VicSee's remote storage, returning a public URL, so an agent could be steered into publishing local media files; the extension allowlist and the declared purpose limit this.

src/index.ts:111

Upload a LOCAL file (image, video, or audio) from this machine and get back a public https URL.

Why it was refuted

Uploading a local file to get a public URL is the tool's declared core purpose, stated in its own metadata. It is limited by an extension allowlist and size caps, and it exposes nothing beyond what the description says.

Data exfiltration

criticalRefutedlow confidence

vicsee_upload reads any agent-supplied local file path (image/video/audio extensions) and PUTs it to a presigned storage URL returning a public URL; this is the declared core purpose, but the tool could upload sensitive local media to a public CDN if an agent is prompted to, with no path restriction.

src/index.ts:131

bytes = await readFile(file_path);

Why it was refuted

The readFile call is the implementation of the declared upload capability. The tool description discloses it, and it sits behind an extension allowlist and size caps.

Data exfiltration

criticalRefutedlow confidence

Any tool taking image inputs (resolveImageInput) reads an arbitrary agent-supplied local path (restricted to image extensions) and inlines it as base64 into API requests to VicSee; the local-file inlining is mentioned only in code comments, not in tool metadata, so it is a lightly disclosed capability that allows any local image file to be sent off-host.

src/index.ts:52

bytes = await readFile(value);

Why it was refuted

Inlining a local image as base64 is a disclosed feature of the wrapper, needed because the API accepts only public URLs or data: URIs. Only image extensions are read, and the data goes to the wrapped VicSee service, which is the declared core function.