Overview
How PostMantis fits into agent, automation, and backend workflows.
One domain, several automation surfaces
PostMantis exposes one publishing model through HTTP, OpenAPI, and MCP. SDKs, wrappers, and skills can sit on top, but they should not redefine the product behavior.
Pick the right surface
HTTP API + OpenAPI
Best for backend services, typed client generation, and direct machine integrations.
MCP
Best for AI clients that need trusted tools for profiles, posts, uploads, and delivery readbacks.
SDKs, wrappers, and skills
Optional convenience layers that should stay thin and preserve the underlying contract.
Recommended flow
Connect profiles in the dashboard
Provider authorization stays in the dashboard. Agents and automations should never perform provider OAuth directly.
Create a scoped API key
Create a dedicated key for the automation. Limit it to the permission level, providers, and profiles it actually needs.
Choose the automation surface
- use HTTP or OpenAPI when you want direct programmatic control
- use MCP when the client is an AI tool or assistant
Accept work, then read results later
Create, schedule, publish, cancel, and delete-draft actions change workflow state. They do not mean provider delivery has already finished.
Observe outcomes cleanly
Read posts later to inspect deliveries[] for the current state of each delivery.
Notes that matter in production
Create and publish are acceptance commands
A successful write means PostMantis accepted the work. Final delivery outcomes appear later.
Profiles come first
List profiles before creating posts when destination IDs are not already known.
Scope stays real
API key permissions, provider scope, and profile scope mean the same thing across HTTP and MCP.
Use readbacks for inspection
Poll posts to inspect delivery state and outcomes.
Common automation patterns
- Draft-first assistant: create drafts for review, then publish only after explicit approval
- Campaign scheduler: create posts with future publish times for one profile or provider set
- Content ops bot: inspect recent posts, failed deliveries, and delivery URLs
- Delivery follow-up: fetch recent posts, summarize outcomes, and flag failures for review
Keep convenience layers thin
PostMantis treats these as the canonical surfaces:
- HTTP API
- OpenAPI
- MCP
Everything else should derive from them:
- generated clients
- thin SDK wrappers
- AI skills and templates
- vendor-specific adapters when truly needed
That keeps the behavior consistent no matter which client is calling PostMantis.