What an agent can do with ShapelessAI
Every capability an agent can drive on a ShapelessAI account, one page each: publish and schedule posts, review the queue, hand over a goal for video and carousels, cast voices, read and write Brand Memory, read performance. Each names the MCP tools, REST calls and CLI lines that do it.
Not a direct call yet
Ask to be told the day one ships: POST https://shapelessai.com/api/capabilities/{id}/want (with a key, or {email} / {webhookUrl}). One with a route answers 501 there today, with the body it will take.
- Render a scripted video in one call (
video-render): Hand the script to a job (POST /api/v1/jobs): the team narrates, scores, captions and renders it per aspect. Route:POST /api/videos/render, {script, aspects?, voices?, music?, captions?} -> {mediaKeys[]}. - Query trends (
trends): The live sensors answer as MCP tools and at POST /api/tools/{name}: x_pulse, news_search, community_hot, related_trends. Route:GET /api/trends, ?platform=&topic=&since= -> {trends[]}. - Find and hire creators (
influencers): Not offered yet: ask at founders@shapelessai.com. Route:GET /api/influencers, ?category=&platform=&budget= -> {creators[]}. - Edit a full-length video (
long-form-edit): edit_footage cuts a take down to a short on its transcript (POST /api/tools/edit_footage). A whole recording back as one edited long-form video, with the shorts cut from it, is the next lane. Route:POST /api/videos/edit-long, {mediaKey, brief?, chapters?, shorts?} -> {mediaKey, chapters[], shorts[]}. - Model content on the creators you want to be like (
identity): Shapeless Graph resolves a creator by handle, pulls their posts and ranks them by multiplier, and finds similar people (POST /api/identity/resolve, GET /api/identity/library?sort=multiplier, GET /api/identity/similar); reference_search answers with the winning videos closest to a brief. One call from handles to formats and outliers is the next lane. Route:GET /api/identities, ?handle=&platform= -> {identity, handles[], outliers[], similar[]}. - Upload a file from a chat host (
upload-from-mcp): The hosted MCP has no disk; the CLI'sassets uploadandPUT /api/brain/assets/{filename}do it.
