Content Distribution MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| DISTRIBUTION_BACKEND | No | State backend (yaml only in v1) | yaml |
| DISTRIBUTION_BACKEND_DIR | No | Directory for YAML state files | ~/.distribution-mcp |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| post_publishA | Publish one or more channel variants immediately. Side effects: makes external HTTP requests to each channel platform; writes publish state to the local YAML backend; requires valid credentials in the named profile. Idempotent on (content.id, channel) — re-running with the same IDs returns cached state without re-posting. Use post_publish for immediate-only delivery; use post_schedule when any variant needs a future schedule_at; use post_drain to flush a previously built queue. |
| post_scheduleA | Enqueue channel variants with schedule_at for future publishing; variants without schedule_at are published immediately. Side effects: writes entries to the local YAML schedule store; makes external HTTP requests for any immediately-published variants; requires credentials in the named profile. Idempotent on (content.id, channel). Use post_schedule when any variant needs a future publish time; use post_publish for all-immediate delivery; use post_drain to process the scheduled queue later. |
| post_drainA | Fire all scheduled posts due at or before the given time boundary. Side effects: makes external HTTP requests for each due entry; writes results to the YAML backend. Idempotent — already-published (content.id, channel) pairs are skipped; no-op when no entries are due. Safe to call from cron. Use post_drain on a recurring schedule to flush the queue; use post_publish or post_schedule to add new content; use post_status to inspect results after drain runs. |
| post_statusA | Return publish state for content pieces. Filters by content_id, channel, or both; returns all entries when neither is given. Side effects: read-only; no external HTTP calls; no auth needed. Deterministic given unchanged backend state. Use post_status to inspect what has been published, what is queued, or what errored; use post_publish, post_schedule, or post_drain to change state. |
| post_unpublishA | Best-effort delete of a published post on the target platform. Side effects: makes an external HTTP DELETE or update request; DEV.to sets published=false (soft delete); platforms without a delete API return success=false without error. Non-idempotent — calling on an already-deleted URL may return a platform 404. Use post_unpublish to retract a live post; use post_status first to obtain the live_url; use post_publish to re-publish after an unpublish. |
| channel_hintsA | Return static per-channel metadata: character limits, Markdown support flags, tag vocabulary, and CTA placement rules. Side effects: read-only; no external HTTP calls; no auth needed. Fully deterministic — returns compile-time adapter constants. Use channel_hints before composing a variant body to understand channel constraints; use post_publish or post_schedule once you have a valid variant. |
| profile_listA | Return all distribution profile names configured in the YAML backend. Side effects: read-only; no external HTTP calls. Deterministic given backend state. Use profile_list to discover available profiles before calling post_publish, post_schedule, or subreddit_list; then pass the chosen name as profile_name. |
| subreddit_listA | Return all subreddits in the Subreddit Catalog with cooldown windows, flair vocabulary, and last-posted metadata. Optionally filtered to subreddits allowed by the named profile. Side effects: read-only; no external HTTP calls. Deterministic given backend state. Use subreddit_list to select a subreddit and obtain flair IDs before composing a reddit: channel variant; pass flair in variant.extras.flair. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 8 tools
Each tool maps to a distinct action or data source, and descriptions explicitly distinguish publish, schedule, drain, status, and unpublish flows. No two tools appear interchangeable.
The five mutating/state tools use a clear post_* prefix, but the read-only metadata tools use noun-first names such as profile_list, subreddit_list, and channel_hints. This creates two naming families, though the overall scheme remains readable and mostly predictable.
Eight tools is well-scoped for a content-distribution workflow covering immediate publishing, scheduling, queue draining, status inspection, unpublishing, and channel/profile discovery. No tool feels redundant.
Core lifecycle is covered: publish, schedule, drain, status, unpublish, plus profile, channel, and subreddit metadata. Notable gap: no explicit cancel or edit operation for scheduled queue entries, and no content-listing tool, though most workflows can work around these.