Documentation

Build with the social layer for AI.

PubMesh exposes one workspace-aware capability surface to browser clients, MCP clients and direct API integrations.

Core model

A PubMesh user belongs to one or more workspaces. Social accounts are connected to a workspace, not to a specific AI client. That lets the same Instagram or LinkedIn connection be operated from PubMesh, ChatGPT, Claude, a CLI, or another authorized agent without reconnecting it each time.

Identity vs. connectors

PubMesh delegates product authentication to LoginMesh. Social providers such as Instagram, LinkedIn, Threads and Facebook remain separate authorizations attached to a workspace.

MCP

The canonical MCP endpoint is https://pubmesh.media/mcp. PubMesh exposes typed tools for integrations, media and publishing. A client discovers tools and invokes only the capability it needs.

server/discover
tools/list
tools/call → pubmesh.integrations.connect
tools/call → pubmesh.media.generate
tools/call → pubmesh.posts.create

HTTP API

HTTP endpoints live under /api/v1. Browser sessions are authenticated by LoginMesh; server-side application data still lives behind the PubMesh data adapter.

GET  /api/v1/health
GET  /api/v1/integrations
POST /api/v1/media/generate
POST /api/v1/posts

Publishing safety

Drafting, scheduling and immediate publishing are distinct operations. PubMesh keeps live publishing explicit so an AI client can prepare reversible work without silently crossing into a public action.