Skip to content
MCP Server Directory logo
media-content

Transloadit MCP Server

Agent-native media processing MCP for video encoding, image manipulation, document conversion, OCR, speech transcription, and other Transloadit operations.

Reviewed overview

MCP 服务器概览

What is Transloadit MCP Server?

Transloadit MCP Server is an agent-native media processing transport (MCP) built on @transloadit/node. It provides streamable stdio and HTTP endpoints to run Transloadit operations like video encoding, image manipulation, document conversion, OCR, and speech transcription either self-hosted or via a hosted endpoint.

How to use the project

  • Install: npm install @transloadit/mcp-server.
  • Run self-hosted (recommended):
    • Stdio (recommended): TRANSLOADIT_KEY=KEY TRANSLOADIT_SECRET=SECRET npx -y @transloadit/mcp-server stdio
    • HTTP: TRANSLOADIT_KEY=KEY TRANSLOADIT_SECRET=SECRET npx -y @transloadit/mcp-server http --host 127.0.0.1 --port 5723 (set TRANSLOADIT_MCP_TOKEN when binding to non-localhost).
    • Docker images available: ghcr.io/transloadit/mcp-server:latest with env vars and port mapping.
  • Hosted endpoint: https://api2.transloadit.com/mcp using bearer tokens minted via npx -y @transloadit/node auth token --aud mcp or the API/SDK.
  • Integrate with agent runtimes (Claude Code, Codex CLI, Gemini CLI, Cursor, OpenCode) by adding the MCP server as a transport (examples provided in README).

Key features

  • Supports video, image, document conversion, OCR, speech transcription and general Transloadit Assemblies.
  • Two transport modes: stdio (embedded agent) and HTTP (local or remote).
  • Self-hosted or hosted (api2) options with configurable auth models.
  • Support for multiple input types: local path, base64, and URL inputs.
  • Resume behavior for assemblies and tus uploads (when assembly_url provided).
  • Prometheus metrics endpoint and configurable metrics auth and path.
  • Tool surface for common operations: create/list assemblies, list robots, list templates, wait for assembly, etc.
  • Docker images, CLI flags, and environment-variable configuration.

Use cases

  1. Embedded agent workflows that need local access to files and CPU/GPU resources for encoding/manipulation via stdio.
  2. Self-hosted media pipelines behind private networks exposing an HTTP MCP with an internal token.
  3. Remote agents that cannot run npx locally using the hosted MCP endpoint and minted bearer tokens.
  4. Automated document OCR and speech transcription pipelines integrated into CI or backend services.
  5. Resumeable uploads and long-running assembly workflows where tus upload continuity is required.

FAQ

  • Can I self-host the MCP server?

Yes. Self-hosting is recommended for agents that need local filesystem access. Run npx -y @transloadit/mcp-server stdio or http modes, or use the provided Docker image.

  • How does authentication work?

Hosted MCP uses bearer tokens minted via npx -y @transloadit/node auth token --aud mcp or API/SDK. Self-hosted stdio/local HTTP requires TRANSLOADIT_KEY + TRANSLOADIT_SECRET; non-localhost HTTP requires a static TRANSLOADIT_MCP_TOKEN you set and send as Authorization: Bearer <token>.

  • What input types are supported and when should I use each?

Supported inputs: path (local files, requires MCP process filesystem access), base64 (small inlined data), and url (remote inputs). For large files prefer path or URL; base64 has hosted limits.

  • Are there limits on inline/base64 requests?

Yes. Hosted default request body limit is 1 MB, hosted maxBase64Bytes is 512,000 decoded bytes. Self-hosted default body limit is 10 MB (configurable).

  • Can MCP resume interrupted uploads?

Yes. If you provide assembly_url, MCP can resume using existing tus_uploads + uploads info; path-based inputs support resume.

  • How do I monitor the server?

Prometheus metrics are exposed at /metrics by default (configurable path and optional basic auth). A public discovery server-card is available at /.well-known/mcp/server-card.json.

  • Where can I find docs and help?

Official docs: https://transloadit.com/docs/sdks/mcp-server/ and API token docs https://transloadit.com/docs/api/token-post/. The repository README includes examples, configuration, and a contributing guide.