PostGun speaks MCP. Point Claude Code, Cursor, or any agent at your workspace and it can read your accounts, generate media, draft, schedule and publish — through the same services the app uses, with a key you scope yourself.
Posts, media, integrations, projects and analytics. Every call validates against real platform limits, so your agent gets "caption too long for X, 312/280" instead of a provider error it cannot read.
02
Publishing is its own permission
posts:publish is a separate scope from posts:write. An agent can compose and schedule all day and still never touch a live account until you grant it, and a key can be pinned to a single project.
03
Dry runs and a full audit trail
Schedule and publish accept dryRun, returning what would happen without writing anything. Every call is logged per key — tool, status, duration — and revoking a key cuts access immediately.
Tool catalog
What an agent can actually do
Each tool declares the scopes it needs. A key only sees the tools its scopes allow, so an agent cannot discover a capability you did not grant.
Posts
create_draft_post
Composes a draft. Never publishes. posts:write
schedule_post
Future date required. Supports dryRun. posts:write
publish_now
Destructive, and gated behind posts:publish
update_post
Drafts and scheduled posts only. posts:write
cancel_scheduled_post
Sends a scheduled post back to draft
list_posts / get_post
Filter by state, date, channel. posts:read
delete_post
Destructive, and gated behind posts:delete
Media and workspace
upload_media_from_url
The video and image ingest path. media:write
list_media
What is already in the workspace. media:read
list_integrations
Connected accounts and disabled state
get_platform_constraints
Caption limits and media rules per platform
list_projects
Projects a key may act inside. projects:read
Orientation
whoami
Org, plan, credits and the scopes this key holds
get_analytics_summary
Performance for the connected channels. analytics:read