Transcodely

What Transcodely MCP server does

The Transcodely MCP server provides an agent-facing video pipeline over Model Context Protocol. An agent can submit a remote HTTPS video for transcoding and hosting in one operation, then retrieve a playable link after processing. It can also create jobs directly, generate AI captions for hosted videos, and inspect the resulting media.

The service is scoped to one Transcodely app per session. Jobs, videos, presets, and ingest rules belonging to another app are not exposed through the tool surface. The server also records an audit trail for tool calls equivalent to the corresponding REST operations.

How it works

Connect to the hosted streamable HTTP endpoint at https://mcp.transcodely.com/mcp. Browser-based clients use OAuth 2.1 with PKCE, and the session determines the app scope. The bare endpoint resolves to the organization’s oldest active app; an app-specific URL can pin the connection to a particular app.

For headless or server-side agents, the connection guide supports sending a Transcodely API key as a bearer token. Three operations start billable work: create_job, create_video_from_url, and generate_captions. Preset creation, reads, reporting, and spend-limit changes do not themselves incur those processing charges.

Setup and configuration

The Transcodely MCP server is hosted, so there is no local service to install or run. In Claude Code, add the HTTP server with:

claude mcp add --transport http transcodely https://mcp.transcodely.com/mcp

Authenticate through the browser when prompted. In Claude desktop, add the same URL as a custom connector. In Cursor, add an HTTP MCP entry pointing to the endpoint in ~/.cursor/mcp.json. Other clients can connect if they support remote MCP servers over streamable HTTP.

The set_spend_limit tool can set or clear a monthly transcoding limit in EUR, but it requires an organization owner or admin using browser OAuth. API-key sessions cannot use this tool.

Tools and capabilities

Available operations include:

• Create transcoding jobs and reusable custom encoding presets.
• Ingest and host an HTTPS video with create_video_from_url.
• Generate captions for a hosted video.
• List jobs, presets, ingest rules, and ingest events.
• Read job progress, per-output status, playback information, and encoded renditions.
• Compare a written output against the job request with get_output_report.
• Review monthly encoding, storage, egress, request, and EUR cost data.
• Update custom presets and configure the monthly spend limit.

Ingest rules describe how storage objects become transcoding jobs. Ingest-event results include the bucket, object key, outcome, reason for skips or failures, and any created job ID.

Limitations and notes

The Transcodely MCP server has no delete, cancel, remove, purge, rotate, or destroy tools. Agents cannot remove videos, jobs, outputs, presets, or rules, and cannot stop queued or running work. It also does not expose API keys, inbound ingest secrets, organization members, plans, invoices, team settings, or other administrative data.

Editing a preset affects jobs created afterward because presets are expanded when a job starts; existing jobs are unchanged. Lowering a spend limit blocks a later job but does not stop work already in progress. The MCP connection itself is free, while processing charges follow Transcodely’s ordinary rates.