SocialOS

Hackathon

Bounty hub for fit, integrations, proof cards, and the fastest route through each submission story.

5active bounties
8live integrations
18proof cards
all-bountiesmode
Canonical bounty hub

This is the single public submission page for all five DoraHacks bounties. SocialOS stays one product, while each bounty section exposes its own framing, integration, live proof, recording route, and deck appendix reference.

Canonical public bounty hub
This page is the judge-facing proof surface. /demo/ and /buddy/ remain auxiliary proof pages, while the interactive product stays localhost-only in the recorded videos.

Watch Final Videos

Start here if you want the recordings first. Each card opens the stable /videos/... watch page, then links back into the matching proof surface.

Start with the final video pack judge-ready

Open the matching bounty recording directly, or use the canonical bounty hub if you want the full proof matrix first.

Open Bounty Hub Open Pitch Deck
AI Agents for Good DoraHacks

Impact workflows with SDG triage, long-term relationship memory, and multi-channel follow-through via Telegram.

Watch final video Proof page

Bounty Map

Start here, jump to the matching bounty section, then open the linked proof JSON if a judge wants structured evidence.

Claw for Human ready

Bring Claw out of the shell and into a guided relationship workspace.

$500 USD Imperial Blockchain 4 proofs live integrated

Explainable relationship OS instead of a generic assistant shell.

Open section Aux page Proof JSON
Animoca Bounty ready

Position SocialOS as an identity-rich agent system for creator, operator, and community workflows.

$1,000 USD Animoca Brands 4 proofs live integrated

Persistent identity and long-horizon memory instead of a one-shot agent demo.

Open section Aux page Proof JSON
Human for Claw ready

Offer a friendship and gratitude coach with guardrails instead of an unrestricted agent.

$500 USD Imperial Blockchain 4 proofs live integrated

Positive social coaching with explicit boundaries and no risky publishing surface.

Open section Aux page Proof JSON
Z.AI General ready

Use GLM in the core SocialOS workflow, not as a side widget.

$4,000 USD Z.AI 10 proofs live integrated glm-5

GLM is a core production path for summaries, reasoning, and content generation, not a checkbox integration.

Open section Aux page Proof JSON
AI Agents for Good ready

Turn SocialOS into a community support and volunteer coordination operating system.

$5,000 USDT FLock.io 9 proofs live integrated qwen3-235b-a22b-instruct-2507

Long-term relationship memory and follow-through for impact work, not a one-turn charity chatbot.

Open section Aux page Proof JSON

Run SocialOS Yourself

Use the public proof site for review, then jump to GitHub if you want to clone the local-first product or wire up optional providers.

Reuse SocialOS locally builder path

The public site is proof-first. The interactive product is reusable on your own machine with quickstart and optional API keys.

View GitHub repo Run locally API setup guide

Live Partner / API Matrix

This is the live integration matrix for the partner APIs and shared product infrastructure that power the five bounty tracks.

OpenClaw Runtime ready live

4 lanes active for orchestrator, memory, compliance, and publishing.

Open
Workspace UI ready live

Judge-facing capture, memory, drafts, queue, and mirror loop.

Open
Buddy Guardrails ready live

Friendship and gratitude mode keeps the experience narrow and trust-first.

Open
Pitch Deck ready live

Public deck now carries the shared story plus bounty-specific appendix slides.

Open
Z.AI GLM ready live

Live GLM routing is active for chat and draft generation using glm-5.

Open glm-5
FLock ready live

Live FLock SDG triage is active using qwen3-235b-a22b-instruct-2507.

Open qwen3-235b-a22b-instruct-2507
Multi-channel Outreach ready live

SocialOS already ships outreach lanes for LinkedIn, X, Instagram, Zhihu, Rednote, WeChat Moments, and WeChat Official Account.

Open
Telegram Volunteer Channel ready live

Telegram bot delivery is ready for volunteer follow-through via 87***46.

Open frelkjsdf_bot

Submission Pack

The repo, deck, public hub, and video pack must tell the same story without drift in wording or proof posture.

1. Pitch / Demo Video5 videos

Record five independent 5-8 minute videos. Keep the SocialOS product backbone fixed, then swap the bounty-specific framing, integration, and live demo section.

2. Public GitHub RepoREADME

Point judges to the shared README, this canonical /hackathon page, and the matching proof JSON for structured evidence.

3. Pitch Deckappendix

The deck stays shared at the top, then adds one appendix slide per bounty in the same order shown on this page.

Bounty Sections

Each section is a complete judge-facing bounty brief with problem, fit, integration, live proof, recording route, and submission assets.

Claw for Human ready live integrated

Explainable relationship OS instead of a generic assistant shell.

$500 USD Imperial Blockchain OpenClaw Runtime 4 proofs

Problem Framing

Use this framing in the first 40-60 seconds of the bounty video.

OpenClaw is powerful for builders, but judges still need to see it translated into a calm, human-readable end-user workflow.

Why SocialOS Fits

Keep the product narrative fixed and only change the bounty lens.

SocialOS already turns OpenClaw-powered agent orchestration into a calm interface with memory, drafts, queue, and reflection.

Partner / API / Infrastructure

This block answers the judge question “how exactly did you integrate the partner or infrastructure?”

PartnerOpenClaw RuntimeImperial Blockchain

OpenClaw runs the orchestration, memory, compliance, and publishing lanes while SocialOS exposes one guided loop across Workspace, Contacts, Drafts, Queue, and Mirror.

Technical ImplementationGET /proofs?bounty=claw-for-human

The Node API persists people, events, drafts, queue state, and mirror evidence in SQLite while the OpenClaw runtime keeps the product lanes separated behind the same UI.

Track Checklistofficial requirements
  • Show a real OpenClaw-powered product instead of a shell-only workflow.
  • Use the shared demo route to make the orchestration legible to non-technical judges.
  • Close on the public hub anchor and proof JSON so the submission stays reviewable after the video.

Live Proof Summary

This block should match the live provider story in the README, deck appendix, and video.

Current proof posture live=true fallbackUsed=false

4 lanes active for orchestrator, memory, compliance, and publishing.

Proof metadata openclaw

Use the proof JSON and public anchor below as the judge-verifiable evidence package.

Recording Route

Each bounty video is independent, but the interaction route stays fixed.

Local recording route/demo

Open /demo → Launch /quick-capture → Show agent trace + evidence cards

Auxiliary public page

Submission Assets

This is Claw translated into a usable relationship operating system, not left as a shell-only interface.

Animoca Bounty ready live integrated

Persistent identity and long-horizon memory instead of a one-shot agent demo.

$1,000 USD Animoca Brands Animoca-style Identity, Memory, and Cognition 4 proofs

Problem Framing

Use this framing in the first 40-60 seconds of the bounty video.

Creator and community workflows break when identity, relationship memory, and long-horizon coordination are not persistent across sessions.

Why SocialOS Fits

Keep the product narrative fixed and only change the bounty lens.

The repo already has persistent people memory, linked identities, event continuity, and explicit agent lanes for memory, compliance, and publishing.

Partner / API / Infrastructure

This block answers the judge question “how exactly did you integrate the partner or infrastructure?”

PartnerAnimoca-style Identity, Memory, and CognitionAnimoca Brands

The same product loop already links Person, Identity, Interaction, Event, Draft, and Mirror state while OpenClaw coordinates the memory, compliance, and publishing jobs needed to keep that identity graph actionable.

Technical ImplementationGET /proofs?bounty=animoca

The schema keeps linked people and event context first-class, and the runtime splits memory, compliance, and publishing responsibilities into visible lanes instead of a monolithic chatbot. That gives SocialOS the identity, memory, and cognition posture Animoca is asking for.

Track Checklistofficial requirements
  • Show identity, memory, and cognition working together across the same person or community thread.
  • Keep the demo focused on long-horizon continuity, not one-off generation.
  • Use the agent-lane trace to make the multi-agent shape explicit to judges.

Live Proof Summary

This block should match the live provider story in the README, deck appendix, and video.

Current proof posture live=true fallbackUsed=false

4 lanes active for orchestrator, memory, compliance, and publishing.

Proof metadata openclaw

Use the proof JSON and public anchor below as the judge-verifiable evidence package.

Recording Route

Each bounty video is independent, but the interaction route stays fixed.

Local recording route/hackathon?bounty=animoca

Open /hackathon → Inspect Animoca card → Jump into Contacts and Studio agents

Auxiliary public page

Submission Assets

This is persistent identity and memory for creator-community operations, not a one-task assistant demo.

Human for Claw ready live integrated

Positive social coaching with explicit boundaries and no risky publishing surface.

$500 USD Imperial Blockchain Buddy Guardrails on OpenClaw 4 proofs

Problem Framing

Use this framing in the first 40-60 seconds of the bounty video.

Agent products are often too open-ended for students, younger users, or trust-sensitive first-time users.

Why SocialOS Fits

Keep the product narrative fixed and only change the bounty lens.

SocialOS already captures people, follow-up, and reflection. Buddy mode narrows that loop to safe, supportive actions.

Partner / API / Infrastructure

This block answers the judge question “how exactly did you integrate the partner or infrastructure?”

PartnerBuddy Guardrails on OpenClawImperial Blockchain

Buddy mode keeps the same SocialOS memory loop but narrows users into four safe tasks and removes any pressure to publish or configure complex automation.

Technical ImplementationGET /proofs?bounty=human-for-claw

The web layer hides risky controls, keeps the product loop loopback-only and dry-run, and routes users into capture, memory, follow-up, and reflection flows instead of open-ended execution.

Track Checklistofficial requirements
  • Show the four constrained Buddy tasks instead of the full operator surface.
  • Make the safety rails visible: dry-run only, trust-first defaults, and simpler language.
  • Keep the route emotionally approachable for younger or first-time users.

Live Proof Summary

This block should match the live provider story in the README, deck appendix, and video.

Current proof posture live=true fallbackUsed=false

4 lanes active for orchestrator, memory, compliance, and publishing.

Proof metadata openclaw

Use the proof JSON and public anchor below as the judge-verifiable evidence package.

Recording Route

Each bounty video is independent, but the interaction route stays fixed.

Local recording route/buddy

Open /buddy → Pick a safe task card → Jump into a prefilled Workspace flow

Auxiliary public page

Submission Assets

Buddy mode is intentionally narrower, safer, and easier to trust than a general-purpose agent.

Z.AI General ready live integrated

GLM is a core production path for summaries, reasoning, and content generation, not a checkbox integration.

$4,000 USD Z.AI Z.AI GLM glm-5 10 proofs

Problem Framing

Use this framing in the first 40-60 seconds of the bounty video.

Production agent workflows often bolt a model onto a demo page instead of making that model a core part of the real product loop.

Why SocialOS Fits

Keep the product narrative fixed and only change the bounty lens.

Workspace support, judge summaries, and draft generation all map naturally onto the current product loop, so GLM can become a first-class provider instead of a decorative plug-in.

Partner / API / Infrastructure

This block answers the judge question “how exactly did you integrate the partner or infrastructure?”

PartnerZ.AI GLMZ.AI

GLM is routed through the native SocialOS flow: judge-facing summary generation, Workspace support, and platform-native draft generation all call the same GLM path instead of a mock adapter.

Technical ImplementationPOST /integrations/glm/generate

The API exposes a live GLM endpoint, records proof metadata into audit evidence, and reuses the same provider-aware routing path inside Workspace and Draft generation so the bounty is part of the product, not a side panel.

Track Checklistofficial requirements
  • Keep GLM as a core product dependency in the live demo and proof JSON.
  • Show the real generation path inside Workspace or Drafts, not an isolated API call only.
  • Close on a runnable prototype, a public repo, and a live proof surface.

Live Proof Summary

This block should match the live provider story in the README, deck appendix, and video.

Current proof posture live=true fallbackUsed=false

2 drafts generated for Community Mentor Follow-up.

Proof metadata glm glm-5

Captured at 2026-03-08T00:47:03.019Z.

Recording Route

Each bounty video is independent, but the interaction route stays fixed.

Local recording route/hackathon?bounty=z-ai-general

Open /hackathon → Call GLM router → Show GLM-tagged Workspace or Draft flow

Auxiliary public page

Submission Assets

GLM is integrated into the real SocialOS loop and captured as live proof, not demonstrated as a decorative plug-in.

AI Agents for Good ready live integrated

Long-term relationship memory and follow-through for impact work, not a one-turn charity chatbot.

$5,000 USDT FLock.io FLock SDG Triage + OpenClaw + Multi-channel Outreach qwen3-235b-a22b-instruct-2507 9 proofs

Problem Framing

Use this framing in the first 40-60 seconds of the bounty video.

Impact workflows usually stop at categorization instead of carrying urgency and next action into a real coordination loop.

Why SocialOS Fits

Keep the product narrative fixed and only change the bounty lens.

Contacts, events, follow-up drafts, and evidence-backed coordination already exist; FLock adds SDG triage and urgency scoring while SocialOS already ships multi-channel outreach lanes.

Partner / API / Infrastructure

This block answers the judge question “how exactly did you integrate the partner or infrastructure?”

PartnerFLock SDG Triage + OpenClaw + Multi-channel OutreachFLock.io

FLock adds live SDG labeling, urgency scoring, and next-step guidance, then SocialOS promotes that result into contacts, events, follow-up drafts, queue-visible coordination, and a Telegram volunteer channel when credentials are present.

Technical ImplementationPOST /integrations/flock/sdg-triage + POST /integrations/telegram/send

The API calls the live FLock structured-model path, marks the route as open-source-model-backed, records provider metadata into proof evidence, and keeps the resulting action inside the same OpenClaw-orchestrated relationship-memory workflow used elsewhere in the product. The impact track also exposes a Telegram send/status/webhook surface plus the existing outreach lanes for multi-channel deployment.

Track Checklistofficial requirements
  • Show OpenClaw, FLock, and the SDG-aligned triage result inside the same workflow.
  • Expose the multi-channel story through outreach lanes and the Telegram volunteer channel.
  • Make the open-source model requirement explicit in the proof metadata.

Live Proof Summary

This block should match the live provider story in the README, deck appendix, and video.

Current proof posture live=true fallbackUsed=false

Live FLock triage and Telegram volunteer handoff are both captured in the public proof pack.

Proof metadata flock qwen3-235b-a22b-instruct-2507

Captured at 2026-03-08T00:47:05.256Z.

Recording Route

Each bounty video is independent, but the interaction route stays fixed.

Local recording route/hackathon?bounty=ai-agents-for-good

Open /hackathon → Run SDG triage → Show channel proof → Promote the result into follow-up actions and drafts

Auxiliary public page

Submission Assets

This is not a charity chatbot. It is impact triage plus relationship memory and actual follow-through.

Proof Cards

Use the JSON links and auxiliary pages only as supporting evidence. The canonical public entry remains this page.

Telegram volunteer channel receive telegram captured

Volunteer mentor check-in received in Telegram

ai-agents-for-good telegram frelkjsdf_bot telegram bot-api openSourceModel=false live fallbackUsed=false
Open page Proof JSON Slide 13 Audit audit_3bb197bf0ffa44518ce89625b1fbeaf7
Telegram volunteer channel send telegram captured

SocialOS AI Agents for Good proof: SDG triage is ready and the next volunteer follow-up is now queued.

ai-agents-for-good telegram frelkjsdf_bot telegram bot-api openSourceModel=false live fallbackUsed=false
Open page Proof JSON Slide 13 Audit audit_c30a41c346f44753b0788fe87e0ffa1a
FLock SDG triage run flock captured

SDG 4: Quality Education · medium · Capture volunteer mentor details, create a workshop event, and generate a structured follow-up plan for students.

ai-agents-for-good flock qwen3-235b-a22b-instruct-2507 web-workspace http openSourceModel=true live fallbackUsed=false
Open page Proof JSON Slide 13 Audit audit_635abbee1b7a463daa92db227d500a6f
Hackathon draft package generated glm captured

2 drafts generated for Community Mentor Follow-up.

z-ai-general glm glm-5 openSourceModel=false live fallbackUsed=false
Open page Proof JSON Slide 12 Audit audit_cf3d95c465564278b2bf59e9404d8777
Workspace hackathon run glm captured

Hi Minghan, great meeting you at the London hackathon. I really enjoyed our conversation about builder communities and the potential for operator dashboards. Would love to stay in…

z-ai-general glm glm-5 openSourceModel=false live fallbackUsed=false
Open page Proof JSON Slide 12 Audit audit_771348c398fd4f5a916230f972ccb38e
Workspace hackathon run glm fallback

I turned Write into a contact draft first. Save it if it looks right, or keep talking and we can refine it together.

z-ai-general local openSourceModel=true fallbackUsed=true
Open page Proof JSON Slide 12 Audit audit_b97ede182f37447188b424f79dde9a75
Workspace hackathon run glm fallback

I turned At The Hackathon into a contact draft first. Save it if it looks right, or keep talking and we can refine it together.

z-ai-general local openSourceModel=true fallbackUsed=true
Open page Proof JSON Slide 12 Audit audit_333344aaf07a4a69bd28c9d2764aefcd
Workspace hackathon run glm fallback

I drafted the contact for you, but the name still needs a quick confirmation. Edit that first, then save it.

z-ai-general local glm-5 openSourceModel=true fallbackUsed=true
Open page Proof JSON Slide 12 Audit audit_404dd23affb445c698905eddc0b3ec3c
GLM generation run glm captured

SocialOS is a production-ready GLM-powered agent system because it combines a robust multi-agent architecture with real-time social media integration, enabling autonomous task exec

z-ai-general glm glm-5 openSourceModel=false live fallbackUsed=false
Open page Proof JSON Slide 12 Audit audit_2a3c58432b2842b3a09c03cfb4620045
OpenClaw runtime powers the product lanes openclaw ready

4 lanes active for orchestrator, memory, compliance, and publishing.

claw-for-humananimocaai-agents-for-good openclaw openSourceModel=false live fallbackUsed=false
Proof JSON Shared appendix socialos/openclaw/runtime.openclaw.json5
Judge-ready Workspace loop already exists ui ready

Quick Capture, Contacts, Drafts, Queue, and Mirror already operate inside one loopback-only product surface.

claw-for-humananimocaz-ai-general workspace-ui openSourceModel=false live fallbackUsed=false
Open page Proof JSON Shared appendix socialos/apps/web/server.mjs
Persistent identity and relationship memory are first-class memory ready

Person, Identity, Interaction, Event, SelfCheckin, Mirror, and MirrorEvidence rows already power recall and follow-up.

animocahuman-for-clawai-agents-for-good memory-graph openSourceModel=false live fallbackUsed=false
Proof JSON Shared appendix infra/db/schema.sql
Buddy mode keeps the experience safe and narrow safety ready

The product stays loopback-only, dry-run by default, and Buddy mode funnels users into four friendly tasks.

human-for-clawclaw-for-human openclaw openSourceModel=false live fallbackUsed=false
Open page Proof JSON Shared appendix socialos/apps/web/server.mjs
GLM router is wired into Workspace and Draft generation glm ready

Live GLM routing is active for chat and draft generation using glm-5.

z-ai-general glm glm-5 openSourceModel=false live fallbackUsed=false
Open page Proof JSON Slide 12 socialos/apps/api/server.mjs
FLock SDG triage is ready for impact workflows flock ready

Live FLock SDG triage is active using qwen3-235b-a22b-instruct-2507.

ai-agents-for-good flock qwen3-235b-a22b-instruct-2507 openSourceModel=true live fallbackUsed=false
Open page Proof JSON Slide 13 socialos/apps/api/server.mjs
SocialOS already ships multi-channel outreach lanes multi-channel ready

SocialOS already ships outreach lanes for LinkedIn, X, Instagram, Zhihu, Rednote, WeChat Moments, and WeChat Official Account.

ai-agents-for-goodz-ai-general social-channels web-workspace + outreach drafts http + publish-package openSourceModel=false live fallbackUsed=false
Proof JSON Shared appendix socialos/apps/api/server.mjs
Telegram volunteer handoff is available for impact follow-through telegram ready

Telegram bot delivery is ready for volunteer follow-through via 87***46.

ai-agents-for-good telegram frelkjsdf_bot telegram bot-api openSourceModel=false live fallbackUsed=false
Open page Proof JSON Slide 13 socialos/apps/api/server.mjs
Deck and public evidence stay submission-ready deck ready

The shared pitch pack stays repo-native, with 4 runtime lanes and public evidence snapshots available for judges.

claw-for-humananimocahuman-for-claw deck openSourceModel=false live fallbackUsed=false
Open page Proof JSON Shared appendix socialos/docs/pitch/VC_DECK_SPEC.md