Discord is the most agent-shaped community platform of the decade. Every server has the same handful of jobs — moderate the chat, answer the same ten questions, post the daily digest, welcome the new joiners — and every server is doing them by hand. The work is repetitive, the volume is real, the cost of a slow response is a member who never came back. In 2026, the agents are good enough to close the gap.

This guide walks through the four community workflows that ship the most value on Discord, the prompts behind each one, the tools that wire them, and a workflow YAML you can run on a schedule. By the end, you have a Discord server where the moderation is on time, the FAQ is instant, the digest posts itself, and the new joiners feel welcome before they have asked their first question.

This is one of the concrete examples of the 4-stage AI workflow pipeline. The shape is the same — trigger, plan, act, review — and the domain is the community surface where open-source projects, indie SaaS, and creator-led businesses all live. The bonus: every active Discord community is a living backlinks channel for the projects that run them. A well-run Discord is a marketing surface disguised as a community.

This article is for the open-source maintainer whose Discord is the front door of the project and is falling behind the questions, the indie SaaS founder whose community is the highest-converting surface they have and is understaffed, the creator whose server is growing faster than the team, and the engineer who is ready to wire the workflow once and stop doing the typing.

Why Discord is the right next agent rollout

Three reasons every community we work with in 2026 puts Discord at the top of the automation list.

The work is real-time, structured, and bounded. Discord has channels, threads, roles, webhooks, slash commands, embeds, and reactions. The data model is structured. The events are typed. An agent that knows the model can read a server the way a senior moderator does, and can act on the same surface with the same fidelity. Most community platforms are not this clean.

The pain is universal and visible. Every Discord server has a #general that is 30% questions the FAQ already answers. Every Discord server has a moderation queue that is six hours behind. Every Discord server has a new joiner who leaves before anyone says hello. These are the visible pain points; they are the visible wins. The team that ships one of these workflows becomes the team that gets to ship the next one.

The marketing angle is real. An active, well-moderated Discord is a backlinks surface, a retention surface, and a top-of-funnel surface. A community where questions get answered in 30 seconds converts new joiners to active members at 3-5x the rate of a community where questions get answered in 6 hours. The agent is the one that makes the 30-second response possible at any scale. This is why the marketing and community teams should be co-owning the Discord automation.

The risk is the same as any other agent rollout: a Discord server with bad automation is worse than a Discord server with no automation, because the bad answer now lives in the place the community trusts. The way to ship this safely is the way we have been shipping every other agent in this series — start small, keep the human in the loop on the high-stakes actions, and measure the things that matter.

The 4 workflows, in order of ROI

We picked these four by combining the same three filters we use for every workflow in this series: routine (every Discord server does them), high-leverage (the savings show up in the team's week and in the community's health), and tractable today (off-the-shelf tools can do them without a custom build). The order is roughly ROI, from highest to lowest, but you should still start with the one that hurts the most on your specific server.

The four:

1. Moderation. Watch every message in real time, classify it (safe, borderline, violation, raid, spam), and act (allow, warn, delete, mute, ban, escalate). The moderation queue goes from "we see it six hours later" to "we see it in six seconds".
2. FAQ auto-reply. Watch every new question in #support, #help, or #ask-anything, match it against the FAQ / docs / knowledge base, and post the answer (or a link to the right doc) within 30 seconds. The top 10 questions stop eating the team's week.
3. Digest bot. Read the day's conversations, summarise the top threads, the unanswered questions, the new members, and the key links, and post a daily digest to #announcements. The members who only check Discord once a day still see what happened.
4. Onboarding. Detect new members, post a welcome message in #welcome, DM a structured onboarding thread, ask the three questions that get them to their first useful action, and follow up 48 hours later if they have not done it. The first-day experience goes from "you got a role and a wave" to "you got a path".

Each one is one OpenClaw agent. They run as a single workflow that lives next to your Discord server. Let us walk through them.

Workflow 1: Moderation

The moderation agent is the one that watches the chat. It does not write replies, does not post digests, does not onboard members. It does one job: catch the bad message before it derails the room, and surface the borderline case to a human moderator with the right context.

The problem: Your Discord server has 8,000 members. 200 messages an hour on a busy day. Two volunteer moderators in different time zones. By the time they see the spam wave, 40 messages are already in the channel. By the time they see the personal attack, the recipient has already left. The agent catches the bad message in 6 seconds.

Inputs:

• A real-time stream of new messages (Discord gateway events)
• The server's rules and code of conduct (a short document in the workspace)
• The server's role and channel structure (read-only from Discord)
• The action policy: what to do for each category (allow, warn, delete, mute, escalate)

Agent prompt:

You are a Discord moderation analyst. Read the new message and the surrounding context (the channel, the author's history in the server, the recent moderation actions). Classify the message into one of five categories and choose the action:

>

- safe — normal conversation. Action: allow. - borderline — uses a slur in a non-targeting way, debates a sensitive topic in good faith, mild profanity. Action: warn (DM the author with a friendly nudge pointing to the rules) and flag for human review. - violation — personal attack, hate speech, doxx, NSFW. Action: delete, mute the author for 24 hours, log the action, and DM the author with the rule they broke and the appeal process. - raid — coordinated spam from multiple new accounts, mass-mention, scam link. Action: delete, ban the author, alert the moderator channel. - spam — single-message spam (self-promo in non-promo channels, off-topic link, "DM me for crypto"). Action: delete, warn the author, log the action.

>

For every action, output a JSON object: {message_id, category, action, rationale, context_summary, confidence}. For borderline cases, also output the message that will be sent to the author. Do not send any messages yourself — write the proposed message to the workspace, and a human moderator approves the action before it executes.

>

Never auto-execute violation or raid actions. Always queue them for human approval. The cost of a false positive (a wrongly-deleted message) is much higher than the cost of a 30-second delay.

>

If the message is in a language you do not understand, set category: borderline and queue for human review. Do not guess.

Tools the agent uses:

• The Discord gateway (websocket subscription to MESSAGE_CREATE events)
• The Discord REST API (read messages, history, member info)
• The workspace's rules and code of conduct (rules.md)
• A small user_history tool that pulls the author's last 10 messages and any prior moderation actions

Output: a moderation queue in the workspace, with the proposed action, the rationale, and the message that will be sent. A human moderator reviews and approves. The agent executes the approved action.

What "good" looks like:

• Median time-to-action under 30 seconds for safe and spam (auto-executed)
• Median time-to-action under 5 minutes for borderline and violation (human-approved)
• 0 false-positive bans per 1,000 messages (the bar is high because a wrongly-banned member is gone)
• 95%+ recall on raid patterns (catching 95% of raids, with 5% false-positive is the right trade)

The trap to avoid: the moderator that is too eager. The agent that auto-deletes every message with mild profanity is the agent that kills the community. The rule above — "always queue borderline and violation for human review" — is the rule that keeps the moderation honest. The community trusts the moderator who explains their action; the community does not trust the moderator who silently deletes.

Cost: roughly 1-2 cents per message classified. At 200 messages an hour on a busy day, that is $5-10 a day, or $150-300 a month. The cost is real; the alternative (a human moderator on 24/7 rota) is much more.

Workflow 2: FAQ auto-reply

The FAQ auto-reply agent is the one that turns "ask a question, wait 4 hours" into "ask a question, get an answer in 30 seconds". It is the highest-ROI workflow on this list for any server with a docs site, a help center, or a known top-10 questions list.

The problem: Your #support channel has 80 new questions a day. 60 of them are answered by the same 5 FAQ entries. The remaining 20 are real and need a human. The team spends 4 hours a day copy-pasting the same 5 answers. The community waits 4 hours for an answer that is already on the docs site. The agent posts the link in 30 seconds.

Inputs:

• A real-time stream of new questions in #support (or #help, or #ask-anything)
• The knowledge base: the project's docs site, the GitHub README, the FAQ, the channel's pinned messages
• The action policy: which channels the agent can post in, which channels require a human to approve before posting

Agent prompt:

You are a Discord community helper. Read the new question in the support channel. Find the best answer in the knowledge base.

>

Steps: 1. Re-write the question as 3-5 search queries 2. Search the docs site and the FAQ 3. Read the top 3 results 4. Compose a 2-4 sentence answer in the customer's language and tone 5. Cite the docs page or FAQ entry by name and link 6. If the answer is not in the KB, or you are not confident, say "I do not have a confident answer for this — would you like me to ping a human?"

>

Match the tone of the channel: technical channel gets a technical answer, general channel gets a friendly answer. Match the customer's language: if they write in Spanish, reply in Spanish.

>

Never invent policies, prices, or features. If the question is about something the docs do not cover, escalate. Do not guess.

>

For the reply, output a JSON object: {reply_text, source_url, confidence, requires_human}. Do not post to Discord yourself — write the proposed reply to the workspace, and a human moderator (or the agent itself, if requires_human: false) posts the reply.

Tools the agent uses:

• The Discord gateway (websocket subscription to MESSAGE_CREATE in the support channel)
• The docs site (HTTP fetch + a small search index)
• The workspace's FAQ (faq.md)
• The Discord REST API to post the reply (after human approval if requires_human: true)

Output: a reply queued in the workspace, with the proposed text, the citation, the confidence. The agent posts low-confidence replies; high-confidence replies are auto-posted. Either way, the action is logged.

What "good" looks like:

• Self-serve rate: 60-80% of questions answered by the agent
• Median time-to-reply under 30 seconds
• CSAT on the agent interaction equal to or higher than the CSAT on the human interaction
• 0 wrong-answer complaints per 1,000 replies (the bar is high because a wrong answer in a public channel is a trust event)

The trap to avoid: the bot that hallucinates. The single biggest failure mode is the agent confidently telling a community member something that is not true ("yes, you can self-host on a Raspberry Pi" when you cannot). The way to avoid this is the rule above — cite the docs, never invent, and escalate on uncertainty. A bot that says "I do not know, let me get a human" is a 4.8-star bot. A bot that hallucinates is a 2.1-star bot. Pick the first one.

Cost: roughly 2-4 cents per bot interaction. At 80 questions a day, that is $1.60-3.20 a day, or $50-100 a month. The alternative (a human responder for every question) is several hours of community manager time a day.

Workflow 3: Digest bot

The digest bot is the one that turns "I missed what happened yesterday" into "here is what happened". It reads the day's conversations, picks the threads that mattered, and posts a structured summary at a time the community is awake.

The problem: Your server has 800 messages a day across 30 channels. The members who check Discord once a day miss the important threads. The members who only check Discord on the weekend miss the entire week. The digest bot posts a 2-minute summary in #announcements at 9am, every day.

Inputs:

• The day's message history (all channels, or a curated list)
• The server's "important channels" list (e.g. #announcements, #releases, #roadmap, #support, #show-and-tell)
• The summary audience: a daily digest for the active community, a weekly digest for the lapsed members
• The post time (e.g. 9am in the server's primary timezone)

Agent prompt:

You are a Discord community digest writer. Read the last 24 hours of conversations in the server. Produce a daily digest that fits in a single Discord embed (max 4,000 characters total). The digest has 5 sections:

>

1. Top threads (max 5). The threads with the most substantive engagement. For each: a 1-sentence summary and a link to the first message. 2. Top questions (max 3). The questions in #support or #help that did NOT get a confident answer from the FAQ bot. These are the questions the team owes a human reply to. 3. New members (max 10). The new members who joined in the last 24 hours. Welcome them by name. 4. Releases and announcements (max 3). Anything posted in #announcements or #releases in the last 24 hours. 5. Show-and-tell (max 3). The best community work in #show-and-tell, #i-built-this, or similar.

>

Do not include: - Off-topic chatter (the meme channel, the "what are you working on" channel) - Spam or deleted messages - Private threads or DMs

>

Append a one-line footer: "Generated by the digest bot. Reply in this thread if something is missing."

Tools the agent uses:

• The Discord REST API (read message history across channels)
• A small engagement_score helper that scores each thread by replies + reactions + unique authors
• The workspace's "important channels" list (digest_config.yaml)
• The Discord REST API to post the digest (with human approval for the first week, auto after)

Output: a Discord embed posted to #announcements, with the five sections, the citations, the formatting. The first week, a human reviews the digest before it posts. After the digest is consistently good, the agent posts it on a cron.

What "good" looks like:

• The digest is read by 20-40% of active members (Discord's own analytics)
• The "top questions" section has a 0% unanswered rate within 24 hours (the team treats the digest as their daily queue)
• The new members section generates 1-3 welcome replies per day (the community uses the digest as a prompt to engage)
• The digest does not become noise — the team throttles if open rate drops below 10%

The trap to avoid: the digest that becomes wallpaper. The community ignores the digest because the digest is always the same shape, always the same length, always the same 5 sections. The way to avoid this is the section list above — it is a default, not a rule. The agent varies the section selection based on what actually happened in the last 24 hours. On a slow day, the digest is 3 sections. On a release day, the digest leads with the release. The digest reflects the day, not the template.

Cost: roughly 5-10 cents per digest. At one digest a day, that is $1.50-3 a month. This is one of the cheapest workflows on this list; the ROI is one of the highest.

Workflow 4: Onboarding

The onboarding agent is the one that turns "you got a role and a wave" into "you got a path". It detects the new member, posts a welcome, DMs a structured onboarding thread, and follows up 48 hours later if the member has not taken the first useful action.

The problem: 40 new members join your Discord every week. They get the @everyone welcome message in #welcome, and then they are on their own. 70% of them leave within 7 days because they do not know where to start. The agent walks them to their first useful action.

Inputs:

• The MEMBER_JOIN event from the Discord gateway
• The server's onboarding template: the three questions, the recommended channels, the first-useful-action definition
• The follow-up policy: 48 hours after join, check if the member has done the action; if not, send a follow-up

Agent prompt:

You are a Discord community onboarding helper. A new member has just joined the server. Send them a welcome DM and a structured onboarding message in #welcome. The onboarding has three steps:

>

1. Welcome DM. A short, warm welcome. Two sentences max. Tell them why the server exists, and tell them you (the bot) are here to help. 2. #welcome channel post. Tag the new member. One sentence: "Welcome, @name! Tell us: what are you building, what are you trying to learn, and what is the first thing we can help you with?" 3. Recommended channels. A short list (max 5) of channels the new member should look at based on their answer to the question in step 2.

>

If the new member does not respond in #welcome within 48 hours, send a follow-up DM: "Hey @name, did you have a chance to introduce yourself? No pressure — just let us know what you are working on and we will point you to the right channels."

>

Do not auto-assign roles based on the response. Do not post in any channel other than #welcome for the new member. Do not send more than 2 messages total in the first 48 hours (the welcome DM and the follow-up if needed).

Tools the agent uses:

• The Discord gateway (MEMBER_JOIN event)
• The Discord REST API to send DMs and post in #welcome
• The workspace's onboarding template (onboarding.yaml)
• A small member_activity helper that checks the member's last 48 hours of activity in the server

Output: a welcome DM, a post in #welcome, and (if needed) a 48-hour follow-up DM. All actions are logged in the workspace for review.

What "good" looks like:

• 60-80% of new members post an introduction in #welcome within 48 hours (vs 20-30% with a generic welcome)
• 30-50% of new members visit at least one recommended channel within 7 days (vs 10-15% with a generic welcome)
• 0 harassment or spam from the bot (the bot's voice is the team's voice; the team owns the bot's tone)

The trap to avoid: the onboarding that feels like a funnel. The new member who gets 5 messages in their first hour — welcome, intro question, recommended channels, role assignment, follow-up — leaves because the onboarding is too aggressive. The rule above — "do not send more than 2 messages in the first 48 hours" — is the rule that keeps the onboarding warm. The agent is a friendly concierge, not a salesperson. Less is more.

Cost: roughly 5-10 cents per new member onboarded. At 40 new members a week, that is $2-4 a week, or $8-16 a month. The alternative (a community manager DMing every new member) does not scale past 10 a week.

The whole pipeline, end-to-end

Let us put all four agents in one OpenClaw workflow so it runs on the Discord gateway and on a schedule.

# /root/.openclaw/workflows/discord-community-pipeline.yaml
name: discord-community-pipeline
trigger:
  kind: gateway
  source: discord
  events:
    - MESSAGE_CREATE
    - MEMBER_JOIN
secrets:
  - DISCORD_BOT_TOKEN
  - OPENAI_API_KEY
  - ANTHROPIC_API_KEY
steps:
  - id: moderation
    agent: discord-moderator
    input:
      message: "{{event.message}}"
      author_history: "{{lookup_history(event.message.author_id)}}"
    output: moderation_queue.json
    on_hold_for_human_review: true   # for borderline and violation

  - id: faq-autoreply
    agent: discord-faq-helper
    when: |
      {{ event.message.channel in channels.support
         && moderation.category != 'violation' }}
    input:
      message: "{{event.message}}"
    output: reply_queue.json
    auto_post_when: "{{reply.confidence > 0.85 && !reply.requires_human}}"

  - id: digest
    agent: discord-digest-writer
    trigger:
      kind: cron
      expr: "0 9 * * *"   # every day 9am in server TZ
    input:
      window_hours: 24
      channels: "{{inputs.digest_channels}}"
    output: digest.md
    on_hold_for_human_review: true   # for the first week

  - id: onboarding
    agent: discord-onboarding
    when: "{{event.type == 'MEMBER_JOIN'}}"
    input:
      member: "{{event.member}}"
      template: onboarding.yaml
    output: onboarding_log.json
    follow_up_after_hours: 48

  - id: notify
    kind: notify
    needs: [moderation, faq-autoreply, digest, onboarding]
    channel: slack
    target: "#community-ops"
    message: |
      Discord pipeline update:
      - Moderation queue: {{steps.moderation.pending}} pending
      - FAQ replies posted: {{steps.faq-autoreply.posted}}
      - Digest: {{steps.digest.status}}
      - Onboarding actions: {{steps.onboarding.completed}}
          

This is the whole pipeline. One YAML, four agents, two event triggers (gateway + cron), and a Slack notification for the community ops team. The pattern is the same as the YouTube pipeline and the customer-support pipeline and the Notion pipeline — bounded agents, inspectable diffs, human-in-the-loop on the high-stakes actions.

Cost and ROI: what this actually looks like

The per-run economics, assuming a Discord server with 8,000 members and 200 messages an hour on a busy day:

Workflow	Cost per run	Runs per month	Monthly cost
Moderation	$0.015	~50,000 (busy hours)	$750
FAQ auto-reply	$0.03	~2,400 (80 questions/day)	$72
Digest	$0.08	30 (daily)	$2.40
Onboarding	$0.08	~160 (40 new members/week)	$13
Total			~$837/month

For a server at this scale, the alternative is a 24/7 human moderator rota (3-4 part-time moderators) plus a community manager spending 4 hours a day on FAQ responses. Fully loaded, that is $8,000-15,000 a month. The four workflows pay for themselves in week one, and they let the human team focus on the 5% of interactions that actually need a person.

For smaller servers (under 1,000 members), the cost is closer to $50-100 a month, and the alternative is the founder doing the work in their off-hours. The pipeline pays for itself in the founder's sleep.

What can go wrong

A few real failure modes, in order of how often they bite.

The moderation agent is too eager. The agent deletes every message with mild profanity, warns every member who debates a sensitive topic, mutes every new joiner who posts in #general. The community feels policed. Fix: the rule — "always queue borderline and violation for human review" — is enforced. The agent's job is to surface, not to decide. The human moderator is the one who decides.

The FAQ bot hallucinates. The agent confidently tells a community member something that is not true. The community loses trust. Fix: the rule — "cite the docs, never invent, escalate on uncertainty" — is enforced. A bot that says "I do not know, let me get a human" is a 4.8-star bot. A bot that hallucinates is a 2.1-star bot. Pick the first one.

The digest becomes wallpaper. The community ignores the digest because the digest is always the same shape. Fix: the digest is a default, not a rule. The agent varies the section selection based on what actually happened. On a release day, the digest leads with the release. On a quiet day, the digest is 3 sections. The digest reflects the day, not the template.

The onboarding feels like a funnel. The new member gets 5 messages in their first hour and leaves. Fix: the rule — "do not send more than 2 messages in the first 48 hours" — is enforced. The agent is a friendly concierge, not a salesperson. Less is more.

A real raid slips through. The agent misses a coordinated raid because the raid is more sophisticated than the patterns the agent has seen. Fix: the agent's rule — "0 false-negative tolerance on raid patterns" — is the right rule, but the metric that matters is the missed-raid rate, not the false-positive rate. The team runs a weekly review of the moderation queue. The agent is not the only line of defense; the team's eyes are the second line.

The pattern across all of these: the agent is a tool, not a replacement. The team is the one that decides when the tool is misbehaving and fixes the prompt, the threshold, or the policy. The agent does not grade its own work.

The marketing angle: why the community team and the growth team should co-own the Discord

The community Discord is one of the highest-ROI marketing surfaces in a developer-led or creator-led business. Three reasons it matters more than most teams realise.

It is a backlinks channel. Every link posted in a Discord channel is a signal to the search engines and the social platforms that the project is alive. A well-moderated, well-trafficked Discord shows up in brand searches, in project comparisons, in "is this project still alive" queries. The open-source projects with the best SEO are the projects with the most active Discords. This is why the marketing team should be co-owning the moderation: every deleted spam message is a small SEO win.

It is a top-of-funnel channel. A community where questions get answered in 30 seconds converts new joiners to active members at 3-5x the rate of a community where questions get answered in 6 hours. The active members become contributors, customers, advocates. The FAQ auto-reply agent is the one that makes the 30-second response possible at scale. The growth team should be co-owning the FAQ agent because the agent is the funnel.

It is a retention channel. The members who feel welcomed and walked to their first useful action in the first 48 hours are the members who are still here in 6 months. The onboarding agent is the one that does the walking. The customer success team should be co-owning the onboarding agent because the agent is the retention surface.

The community team is the steward of the agent. The marketing, growth, and customer success teams are the co-owners. The pipeline is the substrate. The Discord is the surface.

FAQ

What is the best AI bot for Discord in 2026? For the full four-workflow pipeline, OpenClaw with the Discord gateway wired in is the cleanest off-the-shelf option. For one workflow (e.g. moderation), MEE6, Dyno, or Carl-bot with the AI add-on may be enough. For FAQ only, a dedicated tool like AnswerOverflow or a custom GPT integration works. The pipeline is the value; no single tool does the whole thing alone.

Can AI moderate a Discord server? Yes, with the right guardrails. The agent classifies the message and queues the action; a human moderator approves the high-stakes actions. The agent's job is to surface, not to decide. For the safe-and-spam categories, the agent can auto-execute. For borderline and violation, the action waits for a human. This is the right shape — a fully autonomous moderator is too eager, a fully human moderator is too slow.

Is there an AI bot that auto-replies to Discord questions? Yes. OpenClaw, MEE6, and several dedicated tools can read a question in #support, match it against a knowledge base, and post the answer. The trade-off vs a dedicated tool is that OpenClaw gives you the full pipeline (moderation + FAQ + digest + onboarding) in one workflow, while a dedicated tool gives you one workflow and assumes you will wire the rest yourself.

How much does it cost to automate a Discord server with AI? The four-workflow pipeline above costs roughly $837/month for a busy 8,000-member server, and $50-100/month for a small server under 1,000 members. The alternative is a 24/7 human moderator rota plus a community manager, which is $8,000-15,000/month fully loaded at scale.

Can I use this without OpenClaw? Yes, but it is more work. The four workflows can be wired with a Python service using discord.py, or with a low-code tool like n8n or Zapier with the Discord node. The shape is the same. OpenClaw is the substrate that gives you the gateway event trigger, the secret store, the budget, the retries, and the dashboard in one place. The other tools give you the building blocks and assume you will wire the substrate yourself.

Will AI replace community managers? No. The community manager is the person who decides what the community is for, what the rules are, what the tone is. The agent is the person who does the typing and the watching. The two are complementary. The community manager becomes more valuable, not less, because their leverage per hour goes up by 5-10x. The agent is the junior moderator the community manager has always wanted; the community manager is the editor the agent's output needs.

What about existing moderation bots (MEE6, Dyno, Carl-bot)? They are good at the rule-based part of moderation (block this word, mute this role, auto-delete this). They are not good at the content-aware part (this message is a personal attack in a way the rules do not capture). The right shape is to use the existing bots for the rule-based part and OpenClaw for the content-aware part. The two are complementary.

Can the agent handle multiple Discord servers? Yes, with one bot per server and a small dispatch step in the workflow. The agent takes the server id as input and writes to the right one. Most communities have one server, so this is rarely needed, but the pattern is there if you grow into multiple servers (e.g. one per project, one per region).

What is the difference between an AI agent and a Discord bot? A bot answers a command. An agent reads the conversation and acts. The two are at different levels of abstraction. A bot is one of several components in an agent. The pipeline above is an agent that uses the Discord bot framework as one of its tools.

How long does it take to ship the first Discord workflow? A half-day for moderation on one channel with the human-approval gate. A full day for FAQ auto-reply on one channel against a small knowledge base. A half-day for the daily digest. A half-day for the onboarding flow. Two days for all four live, calibrated, and measured.

Will this work for a small server (under 100 members)? Yes, with a smaller cost. The cost-per-message is the same; the volume is lower. For a 100-member server with 50 messages a day, the four workflows cost $5-10 a month. The value is the same — a small server is a small community, but it is still a community that needs moderation, FAQ, and onboarding.

Try it on GolemWorkers

The four agents in this guide, plus the workflow YAML, are available as a one-click template on GolemWorkers. The hosted version wires the Discord gateway, sets up the secret store, and gives you a dashboard tab for the community pipeline. The self-hosted version is the same code, run on your machine. The link is in the description.

If you have a specific server shape (open-source project, indie SaaS, creator community, gaming clan) and a specific scale (small / medium / large), the template adapts. The pipeline is a graph, not a monolith. Bring your own server. Bring your own rules. The four agents are the value.