botverse-mcp

Purpose

Botverse MCP is the MCP server and command-line tool for Botverse, aimed at AI agents and the users who configure them, used to complete video transcoding and document conversion without the need for AWS, FFmpeg, or self-hosted infrastructure.

Main capabilities

• Video transcoding: Supports MP4 (H.264), WebM (VP9), ProRes 422, GIF, and extracting MP3 audio from video. Price is $0.25/task.
• Video/audio conform: Mixes separate video sources and audio sources into one output, reconciling frame rate and duration mismatches (including true pulldown-style speed adaptation, e.g. 24→25fps). Price is $0.30/task.
• Conform splice: Joins video and/or still image segments into one output, such as opening plates, closing plates, inserts/cutaways, or directly merging multiple segments. Price starts at $0.30/task.
• Document conversion: Markdown ↔ DOCX ↔ PDF ↔ HTML ↔ XLSX conversion. Price is $0.05/file.
• Transcription: Transcription with speaker annotation (speaker diarization + AI speaker naming), outputting txt/srt/vtt/docx/pdf. Price is about $3/hour.
• Workflows: Chains multiple Botverse operations (transcoding, conversion, transcription, conform, splice) into a single multi-step task, automatically handling dependency order and parallel branches, with no need to orchestrate each call yourself. Only conversion-type steps are billed against wallet balance ($0.05/step); transcoding/transcription steps are billed by source duration and require auto-recharge to be enabled.

How to use

Two usage paths are provided: the MCP server for AI agents, and the shell-oriented botverse CLI (suitable for evaluation, CI/CD, cron, scripts, and local coding agents).

1. Sign up at botverse.cloud — free trial available: sign-up comes with $1 in credit, no credit card required. A credit card + 2FA are only required on the first top-up (minimum $5). No monthly fee.
2. Get an API key or connector URL from the dashboard.
3. Add to the MCP client configuration (Claude Desktop / Cursor / Windsurf), or use the connector URL (recommended for claude.ai).

Installation command:npx -y botverse-mcp

Command line (botverse): The same package includes the botverse CLI, which reads files from disk and streams them directly to the API (content does not pass through the LLM), making it a fast path for evaluation, automation, and local coding agents. Each task performs upload → poll → download the completed file to -o(default: current directory). Wildcards and multiple --to formats run as a batch.

Sandbox note: The CLI requires outbound network access to botverse.cloud and the Botverse file storage hosts, so it cannot run in sandboxed agent environments where outbound access is restricted by an allowlist (claude.ai / Claude Desktop). In these environments, use MCP tools instead (e.g. convert_content).

Workflows (BWDL)

submit_workflow Chains multiple Botverse tools (transcoding, conversion, transcription, conform, splice) into a single server-side task, for example transcription → clip extraction → delivery, instead of having the agent call and poll one by one.

Workflow definitions use BWDL (Botverse Workflow Definition Language): a JSON object containing workflow_id and steps arrays. Each step has id、tool(any MCP tool name),inputs object, and optional depends_on(array of preceding step ids).

• Steps run in dependency order; steps sharing the same depends_on run in parallel.
• Subsequent steps can use "$.steps.<step_id>.output_key" Reference the previous step's output with "$.params.<name>" Reference workflow-level parameters—no other template syntax is supported.
• Use submit_workflow to submit, use get_workflow_status to poll every 5–10 seconds until the status is COMPLETED, FAILED, PARTIALLY_FAILED, or CANCELLED; if necessary, use cancel_workflow to stop early (only completed steps are billed).
• Billing: only conversion-type workflows are billed against the wallet balance, at $0.05/step. Workflows that include transcoding or transcription steps require auto-recharge to be enabled (billing varies with source duration).

Example—two-step conversion chain (URL → DOCX → PDF): the complete example code is not provided in the materials.

Pricing

• Video transcoding (≤5 minutes): $0.25/task
• Video overtime: +$0.08/minute
• ProRes 422: $0.35/source minute (minimum $0.50)
• Document conversion: $0.05/file
• Credits never expire.

Use cases and limitations

Suitable for scenarios where you need to provide AI agents with video transcoding, document conversion, transcription, and media stitching capabilities, and do not want to build your own AWS or FFmpeg infrastructure. Limitations include: the CLI cannot run in sandbox environments with restricted outbound access; workflows that include transcoding or transcription steps require auto-recharge to be enabled; the workflow template syntax supports only the two reference forms above.