Will this bloat my context window?

Not if you use the CLI: zero tool schemas, JSON output on every command, exit codes 0-7. If you prefer MCP, the full surface is ~16k tokens of schemas — add tools.include to the schedulala block in config.yaml and keep only the posting tools. The 9 show_* widget tools and the attach_media widget path render only on claude.ai, so filtering them out costs you nothing.

Do I need the MCP server, or is the CLI enough?

The CLI is enough, and on a self-hosted agent it is usually better: no schemas in your context, --json on every command, and exit codes your agent can branch on. MCP adds typed tools if you prefer them — one mcp_servers block in config.yaml, trimmed with tools.include.

How does my agent authenticate?

One API key. Set SCHEDULALA_API_KEY in the environment for the CLI (the skill declares it under requires.env), or put the Bearer header on the mcp_servers block in config.yaml. Keys are minted in the Schedulala dashboard, and revoking one there stops every call instantly.

Can I test without posting anything real?

Yes. Mint an sk_test_ key: every create call is simulated end to end — you get a post id and per-platform sandbox results, nothing publishes, and no quota is used. Swap in the sk_live_ key when the flow looks right.

What is the first-comment thing?

Pass firstComment on any post and Schedulala comments on your own post about 15 seconds after it publishes — the standard way to keep links out of the caption. It works on 8 platforms through the same field (LinkedIn, Facebook, Instagram, YouTube, X, Threads, Bluesky, Mastodon), and the post response reports per-platform delivery with the comment URL.

The skill, the CLI, and the MCP server are free and open (MIT). Publishing runs on a Schedulala plan from $29/mo flat — API and MCP access are included on every plan, no separate API pricing. A free key gets 2 connected accounts and 4 lifetime posts to try the loop.

Schedulala for Hermes Agent

Twelve platforms. Zero schema tokens.

Give Hermes a real social-media surface without the context bloat: schedule to 12 platforms, put the link in the first comment, read the analytics back — through a CLI that adds nothing to your context, or one config.yaml block you prune with tools.include.

Start free trial See pricing

One-linerRead https://schedulala.com/SKILL.md and follow it to set up Schedulala.

12 platforms
0 schema tokens via CLI
tools.include-ready
First comments via API
MIT skill
$29/mo flat

01 / INSTALL

Two minutes, either path.

Everything below is copy-paste real — the same commands the hosted skill at schedulala.com/SKILL.md walks an agent through. Keys come from your dashboard; an sk_test_ key dry-runs the whole loop without publishing anything.

The one-liner (skill install)

Read https://schedulala.com/SKILL.md and follow it to set up Schedulala.

Paste this to your agent. It reads the hosted skill and installs itself.

CLI (recommended — zero tool-schema tokens)

npm i -g schedulala
export SCHEDULALA_API_KEY=sk_live_...
schedulala whoami --json

Node 18+, no other dependencies. Exit 0 = connected, and your context window never hears about it.

Remote MCP (~/.hermes/config.yaml)

mcp_servers:
  schedulala:
    url: "https://schedulala.com/api/mcp"
    headers:
      Authorization: "Bearer sk_live_..."

Trim it in the same block: tools.include with the core posting tools — the 9 show_* widget tools and the attach_media widget path are claude.ai-only dead weight here.

02 / THE FLOW

Key, connect, ask.

01
Get a key
Mint an API key in the Schedulala dashboard (Developer view), or have your agent run schedulala init with your email — it prints a code you approve in the browser. sk_test_ keys dry-run everything without publishing.
02
Point Hermes at Schedulala
Five lines in ~/.hermes/config.yaml for MCP — pruned with tools.include — or skip MCP entirely: the CLI does everything over plain HTTPS with --json output and exit codes your agent can branch on.
One-linerRead https://schedulala.com/SKILL.md and follow it to set up Schedulala.
03
Ask for the week
Schedule the week across your accounts, put each blog link in the first comment, and read back yesterday's numbers. Everything lands on the Schedulala calendar where you can edit or cancel before send time.

Start free trial

Works on every Schedulala plan — API, CLI, skill, and MCP are all included.

03 / POSTING

One prompt. A week on the calendar.

Describe the week you want and your agent writes each caption, adapts it per platform, and schedules the whole run - single posts, carousels, threads on X, Threads, Bluesky and Mastodon, or 25 posts in one bulk call.

Every platform's rules are validated before anything is created, and everything lands on the Schedulala calendar where you can edit, move, or cancel it before send time. No scheduling flag means draft - nothing publishes by surprise.

Bulk, threads, queues, drafts - the whole posting surface, from a conversation.

create_postbulk_create_postscreate_threadvalidate_postupdate_post

04 / THE FIRST-COMMENT FIELD

The link goes in the first comment. Automatically.

Anyone monetizing Facebook or LinkedIn traffic knows the play: keep the link out of the caption, drop it as the first comment seconds after the post goes live. Doing that by hand for every post is exactly the kind of work an agent should own — so it's one field on the API.

Pass firstComment on any post and Schedulala delivers it about 15 seconds after publishing — on 8 platforms through the same field, with per-comment targets on threads and per-platform delivery status (including the comment URL) reported back on the post. A failed comment never fails the post.

Eight platforms, one field. Not an Instagram-only bolt-on.

Platform	First comment	Comment limit	Thread targets
LinkedIn		1,500 chars
Facebook		no fixed limit
Instagram		1,000 bytes
YouTube		10,000 chars
X / Twitter		280 chars
Threads		500 chars
Bluesky		300 chars
Mastodon		instance-defined
TikTok		no comment API
Pinterest		no comment API
Telegram		no comment API
Google Business		no comment API

Thread targets: on X, Threads, and Bluesky every post in a chain can carry its own comment ({target, text}). Full field reference: developers/docs/posts.

05 / ANALYTICS

Real numbers, not vibes.

Ask what worked last month and your agent pulls the actual reach, saves, engagement rate, and follower curve - per post, per account, or merged across everything you have connected.

Best posting times come from your own engagement history, not a blog post. YouTube videos come with transcripts and retention curves, which is how a Short becomes next week's posts.

The same agent that scheduled Tuesday can tell you how Monday did.

get_post_analyticsget_follower_growthget_best_posting_timesget_video_retentionget_video_transcript

06 / ENGAGEMENT

Every comment triaged. Rude ones hidden.

Your agent reads the inbox across platforms, drafts replies in your voice, surfaces the ones that need you personally, and hides the ones you would rather nobody saw. Nothing is ever deleted.

Replies and hides are publishing actions, so the skill tells agents to confirm with you first - the inbox never runs away on its own.

Comments, replies, mentions, moderation - handled where the posts came from.

list_commentsreply_to_commenthide_commentlist_mentions

07 / LISTENING

It hears the conversation, too.

Keyword feeds watch public posts on Bluesky and Threads - your product, your niche, your competitors - and your agent can search them, read a thread's context, and reply from your account.

Up to 20 keywords per platform, managed from the same conversation. This is how an agent finds the customer complaining in public before you do.

Social listening wired into the same loop that posts and replies.

search_feedslist_feed_keywordsupdate_feed_keywordslookup_profile

08 / EVERY TOOL, NAMED

Every tool, named.

This is the whole surface your agent gets: 31 tools plus 10 interactive panels. Hermes filters tools in your own config.yaml — tools.include and tools.exclude take glob patterns, and include wins. The full surface is ~16k tokens of schemas; the CLI costs zero. If you connect MCP, start with tools.include on the core posting tools.

Publish and schedule

Describe the week you want and it lands on your calendar, per platform, with a send time you can still change.

create_post
bulk_create_posts
create_thread
update_post
cancel_post
retry_post
list_posts
get_post
validate_post

Analyze

Ask what worked last month and get real numbers, not vibes, straight from the platform APIs.

get_post_analytics
get_follower_growth
get_best_posting_times
get_video_retention
get_video_transcript

Engage

Read the inbox, answer the good comments, hide the ugly ones. Nothing is ever deleted.

list_comments
list_mentions
reply_to_comment
hide_comment

Listen

Search public posts by keyword on Bluesky and Threads, and keep a watch list Claude can read.

search_feeds
lookup_profile
list_feed_keywords
update_feed_keywords

Attach media

Move your photos and video into a post. These tools carry files, they do not create them.

upload_media
list_media
begin_generated_media_upload
append_generated_media_chunk
finish_generated_media_upload

Accounts and usage

See which accounts are connected, connect a new one, and check where you are against your plan.

list_accounts
connect_account
get_usage
get_upgrade_link

Interactive widgets

These do not answer in text. They render a real Schedulala panel inside the Claude conversation: your calendar, your inbox, an analytics chart, a post preview, or a drop zone for the file sitting on your desk.

show_analytics
show_calendar
show_inbox
show_post_preview
show_comparison
show_connections
show_feeds
show_leaderboard
show_hello_widget
attach_media

31 tools · 10 widgets · every one of them live today

09 / PROMPTS

Steal these prompts.

You do not have to learn tool names. Ask in plain English and your agent picks the tools — these are the asks worth keeping in a note.

The Monday batch

Read my drafts folder and schedule the best 5 as posts across bluesky, linkedin and mastodon this week — put each article link in the first comment, not the caption.

create_post
bulk_create_posts

The link-in-first-comment play

Post the announcement to Facebook and LinkedIn at 9am tomorrow with the signup link as the first comment on both.

create_post

Did it land?

Check yesterday's posts — did every first comment get delivered? Give me the comment URLs.

get_post

The dry run

Use my sk_test_ key and rehearse the whole week's schedule — I want to see exactly what would publish before anything real goes out.

validate_post
create_post

Best time, not guess time

When does my audience actually engage? Pull best posting times per account and move this week's schedule onto those slots.

get_best_posting_times
update_post

Photos and video come from you. Ask Claude to attach media and it opens an upload drop zone right in the conversation.

10 / CONTROL

Your agent, on a leash you hold.

Handing an agent your social accounts is a fair thing to be nervous about. Here is exactly what you keep control of, described the way it actually works.

01The key is the leash
Your agent authenticates with an API key you minted. Revoke it in the dashboard and every call stops immediately — scheduled posts and connected accounts are untouched.
- /dashboard?view=developer
02The calendar is the review step
Everything an agent schedules lands on your Schedulala calendar before it publishes. Edit it, move it, or cancel it — nothing is invisible.
- update_post
- cancel_post
03Dry-run keys exist
sk_test_ keys simulate the entire posting loop — post ids, per-platform results, first-comment delivery — without publishing a single thing or using quota.
- sk_test_
04The skill says confirm first
The hosted skill instructs agents to show you what will go out, where, and when before anything publishes. Credentials and payment only ever happen on schedulala.com, never in the chat.
- schedulala.com/SKILL.md

Claude drafts the words and picks the times. You supply the photos and video.

09 / PRICING

MCP costs nothing extra.

The connector is part of the product, not an add-on. Every plan includes it, tool calls are unlimited, and nothing here is metered per post or per token. You pay for how many accounts you manage.

Personal

$29/mo

10 social accounts

One creator, every platform.

All 12 platforms
Unlimited posts
Analytics
Comments and inbox

Unlimited tool calls

Team

$49/mo

25 social accounts

A brand with people in it.

Everything in Personal
Advanced analytics
3 team seats
Priority support

Unlimited tool calls

Agency

$99/mo

100 social accounts

Many clients, one calendar.

Everything in Team
Unlimited team seats
Unlimited workspaces
Dedicated support

Unlimited tool calls

7 days free, $0 today

Seven days free, cancel any time before it ends. See the full pricing table.

10 / FAQ

The questions worth asking.

Not if you use the CLI: zero tool schemas, JSON output on every command, exit codes 0-7. If you prefer MCP, the full surface is ~16k tokens of schemas — add tools.include to the schedulala block in config.yaml and keep only the posting tools. The 9 show_* widget tools and the attach_media widget path render only on claude.ai, so filtering them out costs you nothing.

Point your agent at one URL. Post everywhere.

The hosted skill installs, connects, and teaches the workflow. Your agent can put a week on your calendar tonight.

Start free trial Or start the 7-day trial, $0 today

The one-linerRead https://schedulala.com/SKILL.md and follow it to set up Schedulala.

Twelve platforms. Zero schema tokens.

Two minutes, either path.

Key, connect, ask.

Get a key

Point Hermes at Schedulala

Ask for the week

One prompt. A week on the calendar.

The link goes in the first comment. Automatically.

Real numbers, not vibes.

Every comment triaged. Rude ones hidden.

It hears the conversation, too.

Every tool, named.

Publish and schedule

Analyze

Engage

Listen

Attach media

Accounts and usage

Interactive widgets

Steal these prompts.

Your agent, on a leash you hold.

01The key is the leash

02The calendar is the review step

03Dry-run keys exist

04The skill says confirm first

MCP costs nothing extra.

Personal

Team

Agency

The questions worth asking.

Point your agent at one URL. Post everywhere.