Botverse

io.github.MkTurner74由 GitHub 账号 MkTurner74 发布v1.5.1更新于 Oct 25, 2026

Transcode/conform/splice video, convert docs, transcribe audio for AI agents. Free $1 trial.

已验证STDIO仅桌面文件与存储媒体与设计

概览

AI 生成的概览

Botverse 是一个付费云端媒体与文档服务,以本地 MCP 服务器形式提供,让助手转码视频、整合与拼接片段、转换文档并转写音频。

可以做什么
从公开 URL 或上传文件转码视频为 MP4、WebM、ProRes、GIF 或提取 MP3,并把独立的视频与音频源合成为一个输出。把视频和静态图片片段拼接为片头板、片尾卡、插入片段或连续剪辑。在 Markdown、DOCX、PDF、HTML 和 XLSX 之间转换文档,支持内联内容、公开 URL 或上传文件。提交串联这些操作的多步骤工作流,支持依赖排序与并行分支,并可轮询、取消和查看状态。通过预签名 URL 上传文件、获取签名下载链接并查询钱包余额。
适用场景
适合助手需要在没有本地 FFmpeg 或基础设施的情况下生成或转换媒体与文档,例如把录音转成文字稿、在格式之间转换报告,或把片段与片头片尾组装起来。也适合把多步骤任务作为一个工作流提交。
运行要求
通过 npx 从 npm 包 botverse-mcp 在本地运行。需要 Botverse 账户,并设置环境变量 BOTVERSE_API_KEY 或 BOTVERSE_CONNECTOR_URL 之一。需要访问 botverse.cloud 及其文件存储主机的出站网络。为付费服务,首次充值时需银行卡和两步验证。
安装前请注意
所有操作都从 Botverse 钱包计费:转码、转换、转写和工作流步骤都会产生费用,含转码或转写步骤的工作流要求开启自动充值,可能持续自动扣费。上传的文件和公开 URL 会发送到 Botverse 云端。连接器 URL 内含会话令牌,应视为机密。工作流会在服务端创建和删除任务;取消只停止尚未完成的步骤。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Botverse,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

botverse-mcp

MCP server and command-line tool for Botverse — video transcoding and document conversion for AI agents and the humans who configure them.

[npm]

What it does

  • Video transcoding — MP4 (H.264), WebM (VP9), ProRes 422, GIF, MP3 extraction · $0.25/job
  • Video/audio conform — mux a separate video and audio source into one output, reconciling frame-rate and duration mismatches (including true pulldown-style speed conforms, e.g. 24→25fps) · $0.30/job
  • Conform splice — concatenate video and/or still-image segments into one output: head slates/bumpers, tail slates/end cards, mid-roll inserts/cutaways, or joining clips together · from $0.30/job
  • Document conversion — Markdown ↔ DOCX ↔ PDF ↔ HTML ↔ XLSX · $0.05/file
  • Transcription — speaker-labelled transcripts (diarization + AI speaker naming) → txt/srt/vtt/docx/pdf · ~$3/hour
  • Workflows — chain multiple Botverse operations (transcode, convert, transcribe, conform, splice) into a single multi-step job with automatic dependency ordering and parallel branches, instead of orchestrating each call yourself · convert-only steps run on wallet balance ($0.05/step); transcode/transcribe steps bill by source duration and require auto-refill

Two ways to use it: an MCP server for your AI agents, and a botverse CLI for the shell — evaluation, CI/CD, cron, scripts, and local coding agents. No AWS. No FFmpeg. No infrastructure.

Setup

  1. Sign up at botverse.cloud — free to try: $1 credit on signup, no card required. A card + 2FA are only needed at your first top-up ($5 min). No monthly fees.
  2. Get an API key or connector URL from your dashboard
  3. Add to your MCP client config

Usage

Claude Desktop / Cursor / Windsurf

json
{  "mcpServers": {    "botverse": {      "command": "npx",      "args": ["-y", "botverse-mcp"],      "env": {        "BOTVERSE_API_KEY": "bv_live_..."      }    }  }}

Or with a connector URL (recommended for claude.ai):

json
{  "mcpServers": {    "botverse": {      "command": "npx",      "args": ["-y", "botverse-mcp"],      "env": {        "BOTVERSE_CONNECTOR_URL": "https://botverse.cloud/mcp?token=bv_sess_..."      }    }  }}

Command line (botverse)

The same package ships a botverse CLI for the shell — it reads files from disk and streams them straight to the API (no content goes through an LLM), so it's the fast path for evaluation, automation, and local coding agents.

bash
export BOTVERSE_API_KEY=bv_live_…        # or BOTVERSE_CONNECTOR_URL=…?token=bv_sess_…
npx botverse convert report.md --to pdfnpx botverse convert *.md --to docx,pdf -o ./outnpx botverse transcode clip.mov --to mp4 -o ./outnpx botverse transcribe call.mp4 --to docx --attendees "Sarah Chen,Mike Torres"npx botverse conform video.mp4 audio.wav --to mp4 --method speed_conform --target-framerate 25npx botverse balance

Each job uploads → polls → downloads the finished file to -o (default: current dir). Globs and multiple --to formats run as a batch.

Sandbox note: the CLI needs outbound network to botverse.cloud and Botverse's file storage host, so it does not run inside sandboxed agent environments (claude.ai / Claude Desktop), whose egress is allowlisted. There, use the MCP tools (e.g. convert_content) instead.

Tools (MCP)

ToolDescription
get_upload_urlGet a presigned URL to upload a file for use with the _media/_file/_video (object-key-based) tools
transcode_from_urlTranscode video from a public URL
transcode_videoTranscode an uploaded video file
conform_from_urlMux a separate video + audio source (public URLs) into one output
conform_mediaMux an uploaded video + audio source into one output
conform_splice_from_urlConcatenate video/image segments (public URLs) — slates, bumpers, end cards, inserts, joined clips
conform_splice_mediaConcatenate uploaded video/image segments — same use cases
convert_contentConvert document content inline (up to 4 MB; sandbox-safe)
convert_from_urlConvert a document from a public URL
convert_fileConvert an uploaded document
submit_workflowSubmit a multi-step BWDL workflow chaining several of the above tools together, with dependency ordering and parallel branches
get_workflow_statusPoll a workflow (and its per-step status) until it reaches a terminal state
cancel_workflowCancel an in-progress workflow; already-completed steps are billed, the rest are not
get_job_statusPoll a job until complete
get_download_urlGet the signed download URL
get_wallet_balanceCheck wallet balance

Workflows

submit_workflow chains multiple Botverse tools (transcode, convert, transcribe, conform, splice) into one server-side job — e.g. transcribe → clip extraction → delivery — instead of your agent making each call and polling each one individually.

  • Workflow definitions use BWDL (Botverse Workflow Definition Language): a JSON object with a workflow_id and a steps array. Each step has an id, a tool (any MCP tool name), an inputs object, and an optional depends_on array of prior step ids.
  • Steps run in dependency order; steps that share the same depends_on run in parallel.
  • A later step can reference an earlier step's output with "$.steps.<step_id>.output_key", and a workflow-level parameter with "$.params.<name>" — no other template syntax is supported.
  • Submit with submit_workflow, poll with get_workflow_status every 5–10 seconds until the status is COMPLETED, FAILED, PARTIALLY_FAILED, or CANCELLED, and stop early with cancel_workflow if needed (you're only billed for steps that already completed).
  • Billing: convert-only workflows run on wallet balance at $0.05/step. Workflows containing a transcode or transcribe step require auto-refill to be enabled (billing scales with source duration).

Example — a two-step conversion chain (URL → DOCX → PDF):

json
{  "workflow_id": "md-to-pdf-via-docx",  "steps": [    {      "id": "to_docx",      "tool": "convert_from_url",      "inputs": {        "source_url": "https://example.com/report.md",        "output_format": "docx"      }    },    {      "id": "to_pdf",      "tool": "convert_file",      "depends_on": ["to_docx"],      "inputs": {        "object_key": "$.steps.to_docx.output_key",        "output_format": "pdf"      }    }  ]}

Pricing

  • Video transcode (≤5 min): $0.25/job
  • Video overage: +$0.08/min
  • ProRes 422: $0.35/source-min ($0.50 min)
  • Document conversion: $0.05/file

Credits never expire. Full pricing →

Links

来源:README.md,提交 8ef3f8f

工具

0
SourceWeft 尚未在沙箱中启动过该软件包,因此其工具未知。

版本历史

2
  1. v1.5.1最新Sep 19, 2026
  2. v1.5.0Sep 16, 2026

评分

已满足适用检查项的 50%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通25 分

    请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。

  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分