Skip to content
Back to search
20
MCP http

weftly

becki/weftly

# Weftly Pay-per-job audio/video processing for AI agents. Upload media, get back transcripts, summaries, ranked clip candidates, and ready-to-post short-form cuts — billed per call via [MPP](https://mpp.directory/) (Tempo USDC) or Stripe Checkout for browser MCP hosts. ## Tools - **`transcribe`** — SRT transcript from any audio/video file. _$0.50 audio · $1.00 video_ - **`summarize`** — long-form summary + transcript. _$0.75 audio · $1.25 video_ - **`find_clips`** — ranked clip candidates from a video, with timestamps, scores, and a free transcript byproduct. Optional `query` for needle-in-haystack mode. _$2.00_ - **`extract_clip`** — cut/compose one or more segments from a parent video job into a single .mp4 (with optional time-shifted transcript). _$0.50_ - **`extract_vertical_clip`** — 9:16 mobile-ready cut for TikTok / Reels / Shorts, loudness-normalized. _$0.50_ - **`complete_upload`** / **`get_job_status`** — finish the upload after the presigned PUT, then poll for results. *More coming soon* ## How it works 1. Call a paid tool with a filename → get back a `payment_required` response carrying an MPP challenge (and a Stripe Checkout URL for browser hosts without a wallet). 2. Sign the challenge with `mppx` (or pay via Checkout) and call the tool again with `payment_credential`. 3. Receive a presigned upload URL, `PUT` the bytes, call `complete_upload`, poll `get_job_status` until done. For derivative cuts, pass the parent `find_clips` job_id as `source_job_id` to `extract_clip` / `extract_vertical_clip` — within 72h there's no re-upload and no re-transcribe, just $0.50 per cut. ## Endpoints - MCP: `https://api.weftly.ai/mcp` (Streamable HTTP, JSON-RPC 2.0) - Web UI: `https://chat.weftly.ai` - Discovery: `/.well-known/mpp.json` · `/.well-known/openapi.json` · `/.well-known/cards/{tool}.md` Failed jobs auto-refund. Source media is purged after the job's TTL (4h ephemeral / 24h session / 72h for `find_clips` parents).

Watch this dependency

Get before/after evidence when observed tools, schemas, authentication access, availability or the endpoint changes.

Inspect evidence and set up a watch Share public evidence report
Uptime
0.0%
3 direct probes · 30d
Response
221ms
last probe
Tools
—
not inspected yet
Resources
0
readable
Prompts
0
available
Score: 20/100
Our probe could not complete the MCP handshake.
why this score
fail Answers the MCP handshake 10/50
initialize returned HTTP 405: init HTTP 405.
How to earn +40 points
Return a valid initialize result
POST a JSON-RPC server/discover to the primary URL yourself (2026-07-28: stateless, with _meta and the Mcp-Method header); it must answer 200 with the supported versions, capabilities and serverInfo. Servers on earlier revisions answer initialize with protocolVersion, capabilities and serverInfo.
Docs →
n/a Lists tools, resources or prompts 0/30
Not scored: a catalogue only counts once the handshake succeeds.
n/a 3 or more catalogue items 0/10
Not scored: a catalogue only counts once the handshake succeeds.
pass Describes itself 10/10
1998 characters; more than 30 earns these points.

How to use

Add to your Claude Desktop / Cursor / Cline MCP config:

{
  "mcpServers": {
    "weftly": {
      "url": "https://weftly.ai/mcp/",
      "transport": "http"
    }
  }
}