Integrations
Every door onto a Shapeless account by platform: the MCP server, the Claude Code plugin, the AI SDK tools, the Dify plugin, the Gemini CLI extension, the skill, the CLI and the API.
Everything we publish so an agent platform can reach a Shapeless account without reading the API first. Each one is a thin door onto the same account, the same scopes and the same rail as the API, the CLI and the MCP server, and the source of every one is in the public repo, github.com/FirstClassTree/shapelessai.
Any MCP host
Claude, ChatGPT, Cursor, Codex, VS Code, Gemini CLI, Grok, Perplexity, OpenClaw, Hermes and every
other host that speaks remote MCP: add https://shapelessai.com/mcp, sign in when the tab opens.
One URL, OAuth, no key. The steps for each host, in the host's own words, are on
/connect; the tools are on MCP server.
Claude Code
The plugin wires up the hosted server and ships the skill that teaches Claude the ropes:
Then /mcp, pick shapeless, Authenticate (MCP server).
Vercel AI SDK
shapelessai-ai-sdk on npm: list connected accounts, create a post and read posts as tool()
definitions, with approval before anything goes out. The guide is AI SDK.
Dify
The ShapelessAI tool plugin on the Dify Marketplace carries three tools for workflows, chatflows
and agents: list accounts, create post, list posts. Install it from
marketplace.dify.ai/plugin/shapelessai/shapelessai,
then open Tools, ShapelessAI, Authorize and paste an API key with the read and publish scopes.
Source and README:
integrations/dify.
Gemini CLI
The extension is in the extensions gallery and installs from the repo:
Any agent that reads a skill
SKILL.md is the one file that teaches an agent how to drive the account: first contact, the
shape of the work, the ground rules. It is on skills.sh
and ClawHub, and any agent can fetch it
from shapelessai.com/skill.md.
A terminal, a cron job, CI
The shapeless CLI: npx shapelessai --help. Every command on CLI.
Your own code
The REST API, with an OpenAPI 3.1 document to generate a client from:
API reference, https://shapelessai.com/api/openapi.json.
Where the server is listed
The official MCP Registry
(com.shapelessai/shapeless), Glama
and Smithery. The discovery files on this
domain: /.well-known/ard.json, /.well-known/mcp/server-card.json, /mcp/server.json,
/llms.txt and /llms-full.txt.
Questions
Which integration should an agent builder pick?
The MCP server, unless the platform cannot speak MCP. It carries every tool, OAuth sign-in and the publish gate, and a host learns the surface from the tools themselves. The AI SDK package and the Dify plugin are for a product that embeds posting as one step of its own flow; the CLI is for a terminal or a cron job; the API is for everything else.
Do the integrations share limits with the studio?
Yes. One account, one plan, one rail: a post from Dify counts the same as one from the studio, including the Free plan's ten a day (Posts).