# ShapelessAI [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/FirstClassTree/shapelessai  
**GitHub Stars:** 1  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/shapelessai

## Description
AI social media team for founders: make and publish posts, carousels and short video to 8 networks.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "shapelessai": {
    "command": "npx",
    "args": ["-y","shapelessai"],
    "env": {
      "SHAPELESS_API_KEY": "",
      "SHAPELESS_BASE_URL": ""
    }
  }
}
```

**Requires environment variables:** `SHAPELESS_API_KEY`, `SHAPELESS_BASE_URL` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What ShapelessAI MCP server does

The ShapelessAI MCP server gives an MCP client access to a ShapelessAI account. It covers the account operations exposed by the service, including social posts, durable jobs, standing agents, Brand Memory, assets, connections, and platform rules. Agents can create drafts, schedule or publish content, inspect queue state, and return a studio URL for a job result.

Post operations include creating content for a connection, listing or viewing posts, approving proposed posts, dismissing them, publishing immediately, and recording an externally posted URL. Platform metadata can be read before creating a post so the caller can use the platform’s limits and settings schema.

## How it works

The hosted MCP endpoint is `https://shapelessai.com/mcp`. MCP clients that can open a browser authenticate through OAuth and let the user approve access. Browserless clients can send an API key as a Bearer token to the same endpoint; the key’s scopes determine what the caller may do.

A local mode is available through `shapeless mcp`, which speaks MCP over stdio and uses credentials supplied through the ShapelessAI CLI. Local mode adds disk-related tools such as asset uploads, Brand Memory imports, and files attached to job messages.

Jobs are durable server-side runs. An agent can create a job, inspect its result, watch its event stream, continue a failed or stuck run, or stop it. Job messages can include account media keys and up to six local files. Text and Markdown files are sent inline, while supported media and PDF files are uploaded for the run.

## Setup and configuration

Install the CLI with Node.js 20 or newer using `npx shapelessai` for one-off use or `npm i -g shapelessai` for a persistent command. For local authentication, create an API key under Settings → API keys in the ShapelessAI studio, then run `shapeless login` or provide `SHAPELESS_API_KEY`. Keys can use `read`, `write`, or `publish` scopes; only `publish` permits content to be put out.

The CLI stores its configuration at `~/.config/shapeless/config.json`. `SHAPELESS_BASE_URL` can override the default API host. The hosted MCP connection does not require a local key when OAuth is available. Each CLI command supports `--json` for the raw API response.

## Tools and capabilities

The ShapelessAI MCP server includes tools corresponding to the account and workflow features, including `posts_create`, `jobs_create`, `jobs_continue`, `jobs_list`, `jobs_get`, `jobs_tail`, `posts_approve`, and `brain_write`. Other supported areas include agent management, asset management, connection lookup, platform lookup, and Brand Memory export or import.

The service also exposes the studio agent’s media and research tools as MCP tools. Examples named in the documentation include `generate_image`, `generate_footage`, `edit_footage`, `render_carousel`, `speak`, `caption_file`, `web_search`, and `x_pulse`. Created files are returned as media keys that can be used with post creation. `tools_list` reads the available studio tools and their current details.

Tool annotations identify read-only operations and actions that publish, spend, or overwrite data. Tool descriptions also state the scope they require, allowing a host to request confirmation for destructive actions.

## Limitations and notes

Free accounts are limited to ten posts per UTC day on the publishing rail. Composing, scheduling, and publishing do not consume credits, and the free plan includes a monthly credit allowance for the agent team; paid-plan limits differ as described in the service documentation.

The hosted endpoint depends on an MCP client that supports remote MCP and either OAuth or custom authorization headers. Local file tools are available through the local stdio mode, not solely through the hosted connection. Job event watching is bounded to 60 seconds and 200 events per tail request, and job listing returns 200 items at a time with a cursor for older results.

_Full upstream README: https://allmcps.com/mcp/shapelessai/readme_