Community agent for BDC

[sivangbagri]

1. Objective

• Improve how builders and designers are funneled when they arrive at the BDC website
• Clearly communicate how the community can help and what its current priorities are
• Make it easy to get answers, with a human community fallback when needed
• Build a culture of customer support for the Bitcoin ecosystem, led by the design community
• Routing & Tooling 2026 intiative

---

2. What We’re Building

• An AI agent for the Bitcoin Design Discord community that:
  • Guides new members and helps them get started
  • Provides direction on community initiatives and current priorities
  • Answers general questions and offers assistance anytime

---

3. How We’re Building It

• Using elizaOS, a TypeScript framework for building AI agents that can think, learn, and act autonomously

• Key focus areas:

  • Character design
    • Bio, system prompts, topics, conversation style, example scenarios
  • Context & knowledge
    • Community guidelines, projects, activities, priorities, key people
    • Stored as markdown files or documents
  • Extended functionality via plugins
    • Fetching real-time information on current priorities from external webpages

---

4. Workflow & more

Figma board: https://www.figma.com/board/qNFwGOmJmWN7Y0BGQJHly4/bdc-agent?node-id=0-1&t=E8LnoyOo59hjlJjx-1

[johnsBeharry]

hey there @sivangbagri -- awesome that you’re looking into this and it’s shaping up well. I like how you’ve planned out the technical side in the Figma too.

a few questions that might help push this from a tool to something that improves the expeirnece, and also help communicate how you’re envisioning the bot working in practice. Cause I’m not usually interacting with bots in communities outside of the initial verification bots on telegram (and sometimes mid-journey one), so I’m curious how this actually plays out for a community member day to day.

• Who is this really for?
  A first-time designer landing from the site, a dev with an urgent question, and a returning contributor all probably need different first interactions.

• Where does the agent sit in the journey?
  Is it first-touch onboarding, a Discord-native guide, or more of a fallback when humans aren’t around?

• What should never be automated?
  Which moments must always route to humans (design critique, governance, sensitive conversations)?

• How do priorities stay current?
  Where’s the source of truth, who updates it, and how does the agent reflect changes over time?

• What behaviour are we trying to encourage?
  Saying hello, joining a working group, reviewing work, showing up to calls; not just answering questions.

• How does this lean into Discord itself?
  Roles, Verifications, onboarding questions, channels, guided flows?

might be useful to sketch a simple member journey (arrival -> onboarding -> first action -> deeper participation) and frame the next iteration around that, then layer the AI on top.

Happy to look at a v2 with that lens.

[sivangbagri]

Hii @johnsBeharry
Before moving ahead, I’d like you to go through the flow I added, since it answers some of the questions asked

I’ve prioritized the Ongoing Loops because that’s where real value is created. At the end of the day, the success of our community is determined by active contributors the more we have, the stronger the community becomes.

comparing all the steps, the Ongoing Loop is the one that most directly contributes to success and creates value and actually helps turn a visitor into a contributor. That’s also where the biggest whitespace and challenge exists for both maintainers and new joiners

BDC already has many visitors joining every day but countable active contributors. This agent is meant to be the first real step to solving that helping a visitor reach the point where they are ready to contribute.

Now getting back to your questions :

Who is this really for?

Visitors who want to contribute. Yes the first interaction differs but first interaction is not the best area a where agent can create value

Where does the agent sit in the journey?

Refer Flow

What should never be automated?
Which moments must always route to humans (design critique, governance, sensitive conversations)?

The agent automates structure, not decision-making or authority. Its role is to help community members orient themselves and reach a point where they can start contributing. Beyond that, moments that require judgment such as reviews, critique, taste, governance, decision making are always routed to humans and are never automated.

How do priorities stay current?
Where’s the source of truth, who updates it, and how does the agent reflect changes over time?

Our source of truth is a set of docs and Markdown files (static), plus data exposed by webpages APIs (dynamic). Maintainers can update these, and the agent pulls from them directly, so changes to priorities automatically flow through to what the agent shows and recommends in Discord.

What behaviour are we trying to encourage?
Saying hello, joining a working group, reviewing work, showing up to calls; not just answering questions.

We are trying to encourage contribution by reducing the friction it takes for a member to reach the point where they clearly know what to contribute, where to contribute, and how to do it without relying on manual processes or human intervention.

How does this lean into Discord itself?
Roles, Verifications, onboarding questions, channels, guided flows?

The mentioned discord native stuff are more part of Onboarding and First Action step of user journey not really contributes to Ongoing Loop

[sivangbagri]

Cause I’m not usually interacting with bots in communities outside of the initial verification bots on telegram (and sometimes mid-journey one), so I’m curious how this actually plays out for a community member day to day.

A good example of this is Grok in X, which acts as super intelligent account, which can be tagged anytime, anywhere to ask anything and it doesn't contribute anything in onboarding or verification

[johnsBeharry]

BDC already has many visitors joining every day but countable active contributors.

True, and it's really great this is being tackled. Thanks for this initiative!

The mentioned discord native stuff are more part of Onboarding and First Action step of user journey not really contributes to Ongoing Loop

A good example of this is Grok in X, which acts as super intelligent account, which can be tagged anytime, anywhere to ask anything and it doesn't contribute anything in onboarding or verification

Personally I think they're tied together, one serves the other. For example knowing someone has picked the role of Designer, or Researcher, projects can be filtered out to match them more intelligently. I understand though if you want to keep your scope tight.

Perhaps during a call / community discussion when you present this it can something to get others to handle so you can keep your focus - but I do think it is still quite important to consider.

The example you mentioned sounds great, especially when we use threads. Cause then you can start a thread to brainstorm / while referencing content of the guide, or whatever with someone right in a discord channel without spamming the community.

Will there be memory btw? So will it be able to pick up on past conversations and link people together? I feel like collaboration+connecting with likeminded/aligned interests people in the community is the key for keeping momentum when onboarding.

As a maintainer however much active, it's a big part of what we need to do. Find connection points for collaboration.

For example, these exchanges we're having. If someone feels alone, not responded to, even with the bot, getting assigned tasks - not knowing who has likened interest... you would just feel alone. So finding people who might have shown past interest in a topic like discord bots (there have been), spam prevention, automation, etc that's enough overlap to have easier activation by either joining up, or getting advise/feedback.

---

That's just my quick response, I'll dig in more tomorrow. Have to take sometime away from the keyboard today. Awesome work btw! Love how things are getting fleshed out.

[johnsBeharry]

Just came across a scenario - and I'm not sure how / if it works itself into the bot but thought I'd share.

Someone came into the community asking for help with something related to Saving Satoshi. A Chinese speaker, who needed help, so I tried reaching out to some folks to try to get them a response since I don't know Chinese. I wasn't sure if it was a new contributor to the dev, or just a regular player. I found out that there is something called ₿OSS Challenge (Bitcoin Open Source Software Challenge) and the application process requires people completing the game and sharing their results.

• How might people in the community be aware of such events, and is this a responsibility of the bot to have context of?
• How might we connect people by interest / language in such cases? Is it within scope?
• ... that's all I got for now.

Hopefully this is helpful.

[mouxdesign]

@sivangbagri Thank you for putting this together. Much appreciated!

I had a look at the figma flow but I dont quite understand it upon first glance. I would like to circle back to the points that Johns made in the sense of:

1. When a person arrives at the community, the Bitcoin Design Community is a warm and open and friendly place

• How can we create that warm feel while still using AI?
  Ideas are to use a video format that is a sort of explainer video showing past community initiatives or events

Question: Would the person be able to chat with the bot and ask it questions?

I would gently nudge you to create a community call around this so that we can have more of a free flowing conversation and perhaps you could verbally explain the user flow.

Looking forward to seeing this grow!