FOR AI AGENTS
Give your agent a video editor
Claude Code, Codex, Cursor, Claude and ChatGPT can clip a podcast into Shorts, edit a vlog from raw footage, make a faceless video, caption or dub a video and post it, on your Sentarys account. Connect the MCP server, add the skill so the agent knows the flows, or use the CLI in scripts.
Three ways in
MCP server
126 tools for clips, edits, exports, dubbing, faceless videos, long edits and posting.
MCP referenceAgent skill
Teaches the agent the flows, the credit costs and how to poll long jobs. For Claude Code, Codex and Cursor.
sentarys-skillsCLI
npx sentarys: import a link, upload a file, follow the clips, download them and post. Works in CI.
sentarys-cli
1. Create an API key
- 01Open Settings, API and MCP, and create a key. It starts with snt_live_ and is shown once.
- 02Create one key per agent, so you can revoke one without touching the others.
- 03Keep it in an environment variable: export SENTARYS_API_KEY=snt_live_...
2. Connect the MCP server
Pick your agent and paste the snippet. Claude and ChatGPT use your personal connector URL because they cannot send headers.
Paste a key to fill it into the snippets. It stays in this browser.
Run in your terminal
claude mcp add --transport http sentarys https://api.sentarys.com/v1/mcp \
--header "Authorization: Bearer snt_live_YOUR_KEY"Add --scope user to use it in every project. Then run /mcp in Claude Code to check that Sentarys is connected.
3. Add the skill
The skill tells the agent which tools to call in which order, what each action costs and how to wait for long jobs.
Claude Code
Run inside Claude Code. Installs the skill and the MCP server; the server reads SENTARYS_API_KEY.
/plugin marketplace add Accelar-labs/sentarys-skills
/plugin install sentarys@sentarysWithout plugins, copy the skill folder:
git clone --depth 1 https://github.com/Accelar-labs/sentarys-skills /tmp/sentarys-skills
mkdir -p ~/.claude/skills && cp -r /tmp/sentarys-skills/skills/sentarys ~/.claude/skills/Codex
Run in your project. Codex reads AGENTS.md.
curl -fsSL https://raw.githubusercontent.com/Accelar-labs/sentarys-skills/main/AGENTS.md >> AGENTS.mdCursor
Run in your project. Cursor loads the rule when you ask about videos.
mkdir -p .cursor/rules && curl -fsSL -o .cursor/rules/sentarys.mdc \
https://raw.githubusercontent.com/Accelar-labs/sentarys-skills/main/.cursor/rules/sentarys.mdcCommand line
For scripts, CI and agents without MCP. Log in once; the key is saved in ~/.config/sentarys with access only for you, or read from SENTARYS_API_KEY.
npx sentarys login
npx sentarys credits
npx sentarys import "https://www.youtube.com/watch?v=..." --template podcast --wait
npx sentarys upload ./episode.mp4 --wait
npx sentarys download <projectId> --all --out ./clips
npx sentarys post <projectId> <clipId> --to youtube --caption "New episode" --at 2026-11-01T15:00:00ZAdd --json to any command for machine readable output.
How agents should use it
- Check get_account before a long job and say what it costs: a 40 minute podcast is 40 x 200 = 8,000 credits.
- Long jobs return an id right away. Poll get_project or get_clip every 20 to 60 seconds; never call the paid tool again to check.
- Retries: publish_clip_to_social takes a requestId and remix_video a requestKey. Reuse the same value and nothing is created twice.
- Edits, previews, 1080p exports, downloads and posting are free. Failed jobs refund their credits.
- Ask before posting publicly. For TikTok, the person chooses who can see the post.
Links
- github.com/Accelar-labs/sentarys-skills
- github.com/Accelar-labs/sentarys-cli
- MCP server reference
https://api.sentarys.com/v1/mcp