Hi everyone, I’m Sarah from Nairobi. I want to introduce ZPath, something my team and I have been building for East Africa. It’s a WhatsApp bot called Mwende that helps everyday people get started with Zcash in English/Kiswahili, with no app to download and no account to create. You just chat with it the way you’d text a friend, and it walks you from “what is Zcash?” all the way to your own wallet and your first private transaction. The idea is simple: meet people on the app they already use every day, and make that first step into private money feel easy instead of intimidating.
Application Owners (@octocat, @octocat1)
Organization Name
How did you learn about Zcash Community Grants
Through active involvement in the Zcash ecosystem while building ZPath, our WhatsApp onboarding tool for East Africa. As I worked through Zcash documentation, ZecHub, and the Community Forum, ZCG came up repeatedly as the route for community, education, and ecosystem projects like ours to get support.
Requested Grant Amount (USD)
$25000
Category
Education
Project Lead
Name:Sarah Wahinya
Role:lead
Background:Ecosystem Growth Strategist
Responsibilities:Overseeing the project
Additional Team Members
- Name:Deon Orina
Role:Developer
Background:Rust engineer
Responsibilities:Backend developer
- Name:Samuel Ogera
Role:Developer
Background:Frontend developer
Responsibilities: Developing zpath ui
- Name:
Role:
Background:
Responsibilities:
- Name:
Role:
Background:
Responsibilities:
- Name:
Role:
Background:
Responsibilities:
- Name:
Role:
Background:
Responsibilities:
Project Summary
ZPath is a WhatsApp-native onboarding assistant, Mwende, that guides everyday East Africans from “what is Zcash?” to a created wallet and a first shielded transaction, in English & Kiswahili, with no app to install and no account to create. This grant funds a 3-month Kenya pilot that takes an already-built backend to a live, measured onboarding channel for Zcash.
Project Description
Background
Zcash’s value proposition, privacy-preserving and self-custodial digital money, is exactly what large parts of East Africa need. The region is mobile-first, WhatsApp is the default communication layer, and demand for financial privacy and low-cost value transfer is real. But the on-ramp to Zcash assumes an English-literate, app-installing, technically confident user, and that excludes most of the market.
ZPath closes that gap by meeting people where they already are: WhatsApp. No Play Store, no seed-phrase intimidation up front, no English-only docs. A friendly bilingual assistant (Mwende) walks the user from “what is Zcash?” to a working shielded wallet, in their own language, on the device and app they already use every day.
The backend that powers this, Zbot-core, is already built and running. This grant is about turning a working engine into a measured, production-grade adoption channel for Zcash in East Africa.
What makes this different
It is specifically about Zcash, and specifically for East Africa. Mwende is not a generic crypto chatbot. It introduces users to Zcash, shielded transactions, and self-custody, grounded in indexed Zcash documentation, and routes everything through a privacy-respecting design.
It is already built. This is not a concept. The backend (FastAPI, WhatsApp Cloud API, DeepSeek-powered RAG over ChromaDB with local embeddings, bilingual FAQ engine, admin dashboard, knowledge-base ingestion) is live through Phases 0 to 7. The grant funds productionization, content, and real onboarding, not a cold-start build.
It is privacy-respecting by design. Consistent with Zcash’s values, the system stores only anonymous counters, with no message bodies and no chat logs. The users table holds aggregate counts (total messages, AI queries, FAQ hits, menu taps); AI query logs store language, topic category, latency, and a success flag, never the question text. This is a deliberate architectural choice, documented in the codebase.
Zero-friction reach. WhatsApp has effectively universal penetration in Kenya. Onboarding via WhatsApp removes the single biggest drop-off in crypto adoption: the install-and-create-account wall.
Native bilingual delivery. Every flow ships in English and Kiswahili, with the dispatcher and AI prompt already language-parametrised, and a path to KinyaRwanda, Amharic and French for future regional expansion.
A team that already has the market. The ecosystem lead has spent 4+ years building Web3 adoption across East Africa, with the regulator relationships, founder networks, and community reach to put Zcash in front of real users, not just ship a tool and hope.
Current status (what is already working)
The Zbot-core backend has shipped Phases 0 to 7:
-
WhatsApp Cloud API dispatcher with interactive buttons and list menus, plus a state machine that survives restarts (users are never re-prompted for language).
-
Bilingual (English/Kiswahili) FAQ engine for cheap, deterministic answers to common questions.
-
DeepSeek-powered RAG pipeline with local sentence-transformers embeddings and ChromaDB vector retrieval, grounded in indexed Zcash documents.
-
Admin dashboard wired to real endpoints: KPIs, user counters, FAQ management, knowledge-base upload and ingest, activity log, bot status.
-
Knowledge-base management with PDF, DOCX, TXT, and MD upload, background ingestion, and a rebuild pipeline.
-
Privacy-respecting analytics: anonymous counters and anonymized AI query logs only, with no message content stored.
-
Webhook background-processing so Meta receives a 200 OK in under 100ms even while AI calls run.
Live product surface: https://www.zpath.africa
Backend repository: https://github.com/Zpathafrica/Zbot-core
Content and work plan
This grant covers four workstreams over three months:
-
Production hardening. Take the backend from “works in dev” to “comfortable to deploy”: Alembic migrations, rate-limiting on the public webhook (per-wa_id token bucket), vector-store pre-warm on startup to kill cold-start latency, Redis in production, pagination on admin endpoints, a Dockerfile and compose stack, an automated test suite covering the dispatcher state machine and FAQ engine, and CI on pull requests.
-
Zcash knowledge base. Build and curate a high-quality, Zcash-specific knowledge base (wallet setup for Zashi and Nighthawk, shielded versus transparent transactions, security and self-custody, fees, recovery) in English and Kiswahili, so Mwende’s answers are accurate, grounded, and genuinely useful for first-time users.
-
Live onboarding cohorts. Run real users through the bot in structured cohorts across Kenya (community sessions, university groups, and partner-channel referrals), measuring how many reach a created wallet and a first shielded transaction. Iterate the flows based on where real users drop off.
-
Measurement and transparency. Publish bi-weekly progress updates to the Zcash Community Forum and a final report with full, honest adoption metrics, so the community can see exactly what its funding produced.
Community engagement
-
Bi-weekly progress updates posted to the Zcash Community Forum grant thread, including onboarding metrics, product status, and any timeline adjustments.
-
Open to community feedback on knowledge-base content, onboarding flows, and target user segments.
-
Happy to coordinate with ZecHub and other ecosystem content and education teams on shared material and cross-promotion.
-
All non-personal educational content (flows, knowledge-base material, localized explainers) freely shareable for non-commercial Zcash educational purposes.
-
The backend is open-source; any new open-source components will ship with a CONTRIBUTING.md following librustzcash conventions.
Sustainability
This pilot is designed to prove a repeatable, low-cost onboarding channel. The infrastructure is already lean: the AI stack uses local embeddings (no per-embedding cost) and an OpenAI-compatible LLM, and the only recurring costs are hosting, the LLM API, and WhatsApp messaging. Once the per-user cost-to-onboard is established in the pilot, ZPath will pursue ecosystem partnerships, a follow-up grant for regional expansion (Uganda, Tanzania, Rwanda), and aligned sponsorship to reduce dependency on grant funding. Because the engine is built and open-source, future expansion is incremental, not a rebuild.
Proposed Problem
Zcash has an adoption and accessibility gap in East Africa. The technology is exactly suited to the region’s need for private, low-cost, self-custodial money, but the on-ramp assumes an English-speaking, app-installing, technically confident user, which excludes the mainstream market.
Specifically, East African users face:
-
An install wall. Downloading a wallet app and creating an account is the single biggest drop-off point in crypto onboarding. Many users never get past it.
-
A language wall. Most Zcash educational material is English-only; the mainstream market is bilingual (English/Kiswahili) or Kiswahili-first.
-
A trust and comprehension wall. Seed phrases, shielded addresses, and self-custody are intimidating without a patient, conversational guide in plain language.
-
No native presence on the channel people actually use. WhatsApp is the default communication layer in Kenya; Zcash has effectively no presence there.
The result: a market that needs financial privacy has almost no realistic path to its first shielded transaction.
Proposed Solution
A WhatsApp-native onboarding assistant (Mwende) that takes a complete newcomer from “what is Zcash?” to a created wallet and a first shielded transaction, in English or Kiswahili, with no app install and no account creation, on the channel they already use.
The solution combines a working, privacy-respecting backend (already built) with a curated Zcash knowledge base, real onboarding cohorts in Kenya, and transparent public reporting. ZCG funds productionization, content, onboarding execution, and measurement, not a cold-start build.
Solution Format
Software, plus live onboarding cohorts and public reporting.
-
Production-hardened WhatsApp onboarding bot (Mwende) live on the Meta WhatsApp Cloud API.
-
Bilingual (English/Kiswahili) onboarding flows: language pick → interactive menu → FAQ or free-form AI answer → guided wallet setup → first shielded transaction.
-
Curated, Zcash-specific knowledge base grounding all AI answers (RAG over ChromaDB with local embeddings).
-
Admin dashboard for KPIs, knowledge-base management, and privacy-respecting analytics.
-
Live onboarding cohorts in Kenya with measured conversion to wallet-created and first-transaction.
-
Bi-weekly forum updates and a final public metrics report.
Dependencies
-
Meta WhatsApp Cloud API: the bot runs on Meta’s platform and is subject to their messaging policies and approval. Mitigation: the integration is already built and tested against the Cloud API.
-
LLM provider (DeepSeek, OpenAI-compatible): used for free-form answers. Mitigation: the orchestrator uses the OpenAI-compatible SDK with a base_url override, so the provider is swappable; embeddings are local and provider-independent.
-
Wallet apps (Zashi, Nighthawk): onboarding guides users into existing Zcash wallets. No dependency on those teams for delivery; we link to and document their published flows.
-
No dependencies on ZF, ECC, or any other organization for delivery. This is an independent project.
Technical Approach
Architecture (already implemented)
-
Web framework: FastAPI and Uvicorn.
-
Messaging: Meta WhatsApp Cloud API, with interactive quick-reply buttons and list messages, and webhook background-processing for sub-100ms acknowledgements.
-
Routing: a cheap static-FAQ layer first, then fallback to a DeepSeek-powered RAG pipeline for free-form questions. A state machine (new → awaiting_lang → menu ⇄ ai_chat) with DB fallback ensures users are never re-prompted for language after a restart.
-
AI and RAG: DeepSeek via the OpenAI-compatible SDK; local sentence-transformers (all-MiniLM-L6-v2) embeddings (no extra API key, no per-embedding cost); ChromaDB local vector store grounded in indexed Zcash documents.
-
Persistence: SQLite via SQLAlchemy 2.0 (Postgres-ready).
-
Cache: Redis (preferred), then DiskCache (default fallback), then in-memory, resolved once at startup.
-
Admin: bcrypt-hashed passwords and opaque bearer tokens; full admin REST API consumed by a separate dashboard.
-
Privacy: message bodies are never stored. Only anonymous counters and anonymized AI query metadata are persisted.
Work to be delivered under this grant
-
Pre-pilot and hardening: Alembic migrations; webhook rate-limiting; vector-store pre-warm; production Redis; admin-endpoint pagination; Dockerfile and compose; pytest coverage of the dispatcher and FAQ engine; CI on PRs.
-
Knowledge base: curate and ingest a Zcash-specific, bilingual knowledge base; QA answer accuracy against a fixed question set.
-
Onboarding execution: run structured cohorts in Kenya; instrument and measure conversion to wallet-created and first shielded transaction; iterate flows.
-
Reporting: bi-weekly forum updates; final public metrics report.
Upstream Merge Opportunities
ZPath does not fork or modify any upstream Zcash protocol repositories. It is an adoption and education project built on top of Zcash, guiding users into existing, published wallets such as Zashi and Nighthawk, and is not a change to Zcash core, so no upstream protocol changes are proposed.
-
Repositories forked or modified: none at the protocol layer.
-
Changes planned: none to upstream Zcash software.
-
Benefit to the wider ecosystem: the ZPath backend (Zbot-core) is itself open-source. Any reusable components, for example a WhatsApp onboarding template for other regions and languages, will be published open-source with a librustzcash-style CONTRIBUTING.md so other communities can fork and adapt them.
-
Coordination needed with upstream maintainers: none for delivery. We are happy to coordinate with ZecHub and other ecosystem education teams on shared content.
-
Timeline considerations: not applicable, as there are no upstream merges planned.
Hardware/Software Costs (USD)
$1500
Hardware/Software Justification
Cloud hosting, managed Redis, and database for a production deployment, plus the LLM API (DeepSeek) for free-form answers during the 3-month pilot. Embeddings run locally, so there is no per-embedding cost. This line covers roughly three months of infrastructure and AI usage at pilot volume.
| Item | 3-month cost |
|---|---|
| Cloud hosting (app + worker) | $600 |
| Managed Redis + Postgres | $400 |
| LLM API usage (DeepSeek, pilot volume) | $500 |
| Total | $1,500 |
Service Costs (USD)
$0
Service Costs Justification
No external service contractors are engaged for this pilot. All design, engineering, content, and localization work is performed by the named team and is accounted for under Compensation Costs.
Compensation Costs (USD)
$23500
Compensation Costs Justification
Compensation for the three-person core team across a 3-month pilot. Rates reflect East African market norms and are deliberately lean.
| Role | Scope | Rate | Hours | Cost |
|---|---|---|---|---|
| Project Lead, Sarah Wahinya | Project ownership, ecosystem partnerships, go-to-market, onboarding cohorts, regulatory framing, forum reporting, milestone delivery | $40/hr | 250 | $10,000 |
| Technical Lead, Deon | Production hardening, WhatsApp and RAG pipeline, KB tooling, security, deployment | $45/hr | 200 | $9,000 |
| Product and Localization, Sam Ogera | Content production, bilingual (EN/SW) KB curation, QA, user-testing coordination, community support | $30/hr | 150 | $4,500 |
| Total | 600 | $23,500 |
Total Budget (USD)
$25000
Previous Funding
No
Previous Funding Details
No response
Other Funding Sources
No
Other Funding Sources Details
No response
Implementation Risks
WhatsApp and Meta platform risk. Messaging policy changes or template approval delays could affect delivery. Mitigation: the integration is already built and tested; flows use standard interactive message types; we monitor Meta policy and keep messaging within onboarding and education use.
Lower-than-expected onboarding conversion. Real users may drop off before creating a wallet or making a first transaction. Mitigation: the pilot is explicitly designed to measure and reduce drop-off; flows are iterated cohort by cohort; even partial-funnel learnings are a public deliverable that benefits any future Zcash onboarding effort in the region.
AI answer quality. RAG answers are only as good as the knowledge base. Mitigation: a curated, Zcash-specific knowledge base; a cheap deterministic FAQ layer for the most common questions; QA against a fixed question set; confidence-gated answers with safe fallbacks.
LLM provider dependency and cost drift. Mitigation: the provider is swappable via the OpenAI-compatible SDK; embeddings are local; pilot volume keeps spend bounded and within the infrastructure line.
Timeline slippage. Mitigation: the backend is already built (Phases 0 to 7), which removes cold-start risk; milestones are scoped conservatively; bi-weekly forum updates surface any slip proactively, not after the fact.
Potential Side Effects
None anticipated. All content is educational and factual. The system is privacy-respecting by design, with no message content stored and only anonymous aggregate counters kept, which is consistent with Zcash’s values. Onboarding guides users toward self-custody and shielded transactions using existing, reputable Zcash wallets.
Success Metrics
| Metric | Target | Measurement |
|---|---|---|
| Production deployment | Bot live and hardened (rate-limited, migrated, CI, Dockerized) | Public deployment and repo |
| Bilingual knowledge base | Curated EN and SW Zcash KB ingested and QA’d | Admin KB stats and QA question set |
| Users onboarded (reached menu) | 1,000+ unique WhatsApp users | Anonymous user counters (admin metrics) |
| Wallets created | 300+ users guided to a created wallet | onboarding_stage = wallet_created counter |
| First shielded transactions | 100+ users reach a first transaction | Self-reported completion and funnel instrumentation |
| Answer quality | At least 90% accuracy on a fixed Zcash question set | Manual QA against KB |
| Transparency | Bi-weekly forum updates and a final public metrics report | Forum thread |
All user metrics are aggregate and anonymous. No message content or personal data is stored or reported.
Startup Funding (USD)
$0
Startup Funding Justification
No startup funding is required before milestone work begins. The backend is already built, and the team will begin production-hardening and knowledge-base work before any milestone payment is made.
Milestone Details
- Milestone: 1
Amount (USD): 9000
Expected Completion Date: 2026-07-31
User Stories:
- "As a first-time user, I want a reliable, fast WhatsApp bot that does not break, so that I can learn about Zcash without friction."
- "As a Kiswahili-speaking newcomer, I want accurate Zcash answers in my own language, so that I can understand the technology."
Deliverables:
- Production-hardened backend deployed: Alembic migrations, webhook rate-limiting, vector-store pre-warm, production Redis, admin pagination, Dockerfile and compose, pytest coverage of the dispatcher and FAQ engine, CI on PRs.
- Curated, bilingual (English/Kiswahili) Zcash knowledge base ingested and QA'd against a fixed question set (at least 90% accuracy).
- CONTRIBUTING.md (librustzcash-style) published for the open-source backend.
- Bi-weekly progress updates posted to the Zcash Community Forum.
Acceptance Criteria: Bot is publicly live and hardened; knowledge base is ingested and passing QA; repo and deployment links shared on the forum grant thread. Payment requested only after deliverables are publicly verifiable.
- Milestone: 2
Amount (USD): 9000
Expected Completion Date: 2026-09-30
User Stories:
- "As an East African newcomer, I want to be guided from zero to a created Zcash wallet over WhatsApp, so that I can hold private money without installing a complex app first."
- "As a community member, I want shareable, bilingual onboarding I can send to friends and family, so that I can bring others into Zcash."
Deliverables:
- At least 1,000 unique users onboarded (reached menu); at least 300 guided to a created wallet.
- Structured onboarding cohorts run across Kenya (community sessions, university groups, partner referrals).
- Funnel instrumentation and iteration based on real drop-off data.
- Bi-weekly progress updates posted to the Zcash Community Forum, including anonymous onboarding metrics.
Acceptance Criteria: Onboarding targets met and evidenced via anonymous admin metrics; cohort summary posted to the forum grant thread. Payment requested only after metrics are reported.
- Milestone: 3
Amount (USD): 7000
Expected Completion Date: 2026-10-31
User Stories:
- "As a newly onboarded user, I want to complete my first shielded transaction, so that I have actually used Zcash, not just learned about it."
- "As the Zcash community, I want a transparent report on what this funding produced, so that I can evaluate impact and a path to regional expansion."
Deliverables:
- At least 100 users reach a first shielded transaction.
- Final public metrics report (users onboarded, wallets created, first transactions, language split, answer-quality QA, per-user cost-to-onboard) posted to the Zcash Community Forum.
- Regional expansion plan (Uganda, Tanzania, Rwanda) and a sustainability and partnership plan.
Acceptance Criteria: First-transaction target met and evidenced; final analytics and expansion report posted to the forum grant thread. Payment requested only after the report is public.
Supporting Documents
- ZPath product site (https://www.zpath.africa): live WhatsApp onboarding bot landing page and product surface.
- Zbot-core backend repository (https://github.com/Zpathafrica/Zbot-core): open-source FastAPI backend powering Mwende, including the dispatcher, RAG pipeline, and admin API.