A territorial AI agent with voice interaction and visual elements, built in partnership with MuCA (Museu de Cultura Amazônica) and launching publicly in April 2026. Six chat modes, six non-negotiable principles, zero folklorization. A case of AI serving humans, never above them.
In short. Muiraquitãna is a territorial AI that speaks for a place (Baixo Tapajós), not just about it. Public launch in April 2026 in partnership with MuCA (Museu de Cultura Amazônica). Six chat modes, six principles, zero folklorization. With voice and visual elements. Community first. Territorial.
Generative AI arrived in the Amazon as quickly as it did everywhere else, but the region's cultural, ecological, and social weight does not fit into a generic chatbot. Most tourism assistants, booking bots, and "Amazon-themed" avatars folklorize the territory: reducing a living culture to exotic vignettes, in English, optimized for outsiders.
Baixo Tapajós, Santarém, Belterra, Alter do Chão, FLONA Tapajós, and the riverside and Indigenous communities along the Tapajós and Arapiuns rivers deserved the opposite: a rooted, careful, and useful assistant that serves residents first and visitors second; understands the oral cadence of Portuguese and territorial references; and never competes with the museum, guides, or community voices it supports.
We spent late 2025 and the first quarter of 2026 running an editorial stress test with residents, educators, and researchers (closed sessions, structured feedback, iteration), preparing Muiraquitãna for its true milestone: a public launch in April 2026, with voice and visual modes ready for the territory.
MuCA (Museu de Cultura Amazônica) is a cultural and research center for Baixo Tapajós, based in Belterra (Pará). Data Riders partnered with MuCA as an AI partner, co-designing an agent that acts as a curator of the territory rather than a generic chatbot. The partnership is supported by:
- A public institutional agreement recognizing MuCA as the curatorial authority over the collection and territorial content.
- A joint AI Education Project providing prompt engineering and AI literacy to local residents, educators, and researchers.
- Partnerships with universities, research institutes, and local entrepreneurship support organizations for content validation and referral workflows.
Chat only
No avatar, no artificially generated voice, no synthetic face. Muiraquitãna is text, ensuring it cannot impersonate a person from the territory.
Voice & visual
Native voice interaction in Brazilian Portuguese, with rich visual elements, photos, maps, and collection images designed for the Tapajós context. English and Spanish are secondary.
Territorial tone
The scope is explicit: Santarém, Belterra, Alter do Chão, FLONA Tapajós, and riverside communities along the Tapajós and Arapiuns rivers. It states this clearly, refusing to speak for what is not its own.
No folklorization
No ornamental mysticism, no exotic clichés. Ritual content receives factual captions and community consent.
Community first
Recommendations prioritize local guides, inns, studios, and markets, without displacing local businesses for platform brands.
Curatorial trail
Every sensitive response is traceable to an editorially accountable curator, collection item, or community contribution.
Muiraquitãna was designed as a single voice with six explicit interaction modes. The agent switches modes based on user intent; it never pretends to be multiple personas.
- Territorial Host, welcomes visitors, introduces points of interest and the rhythm of the territory.
- Living Agenda, events, ceremonies, classes, fairs, and gatherings for this week or weekend.
- Services Guide, local guides, inns, restaurants, studios, transportation. Curated, community-first listings.
- Collection Guardian, access to the collective collection of MuCA: images, sounds, stories, factual captions for rituals.
- Community Copilot, helps residents submit events, services, or collection items for curatorial review.
- Scientific Mediator, bridges researchers and the territory: finds papers, summarizes them, explains their relevance to Tapajós.
Data Riders served as the AI product lead, from editorial design to deployment. The work combined cultural research, system prompt engineering, safety work, and product operations:
Editorial design
We wrote the voice guidelines, the six-mode protocol, refusal patterns (out-of-scope requests), and the style guide for ritual captions, in collaboration with MuCA curators.
Safety engineering
Anti-folklorization filters, refusal patterns for medical/legal/financial advice, geofencing for territorial claims, and community-first ranking.
Product operations
Hosting, privacy-preserving telemetry, curatorial inbox for the Community Copilot mode, and an open channel with MuCA curators to refine the voice over time.
Muiraquitãna runs on the Data Riders platform, embedded on the Data Riders desktop site and open in a dedicated window on mobile. The stack is intentionally simple and robust: a curated knowledge base of territorial content (MuCA collection, institutional partners, community submissions), a lean system prompt implementing the six principles and six modes, a retrieval layer biased toward community-first sources, and an editorial workflow (curatorial inbox + weekly review) that keeps the agent active without drifting.
- Key milestone. Muiraquitãna's public release takes place in April 2026, with all six chat modes, native voice interaction in Brazilian Portuguese, and rich visual elements (photos, maps, collection images) available to residents and visitors.
- Location. MuCA, Belterra (Pará), 43 km south of Santarém, on the banks of the Tapajós, plus remote access through the Data Riders website.
- What goes live in April 2026. The full six-mode protocol, voice + visual interaction, the Community Copilot inbox, and the AI Education Project PDF distributed to Baixo Tapajós educators.
- Primary audience. Residents first, followed by researchers, educators, cultural institutions, and visitors, in that order.
Reach
Thousands of sessions in the first weeks, with a growing share of returning residents, not just passing visitors.
Service Guide recommendations are dominated by local guides, pousadas, and artisan studios, not platform brands. Measured weekly.
Curatorial inbox
Dozens of community-submitted events, services, and collection items processed by Copilota mode in the first cycle.
AI Education Project
Educators and students in Baixo Tapajós received the curriculum PDF; classroom sessions began in Santarém and Belterra.
Collection activation
Guardiã do Acervo mode surfaced MuCA material that was previously invisible, now visible with factual captions and context.
Refusal discipline
Out-of-scope requests (medical/legal/financial, other regions) are declined with a warm, localized standard, without hallucinations.
Muiraquitãna is Data Riders' public argument against the AI-first slogan. AI was useful here precisely because it was subordinated to an editorial protocol written by humans with cultural authority. The agent does not try to replace guides, educators, or curators; it directs people to them. It does not speak for the territory, it speaks from within it, and it stops when it does not know. This is the same governance approach Data Riders applies to mining: tools in service of technical judgment, never above it.
This case study deliberately omits personal identifiers and individual community members. Institutional partners (MuCA, universities, research institutes, local entrepreneurship support organizations) are credited by role rather than name in public marketing, per our anonymization policy. For a more detailed editorial briefing, contact us.
We help cultural institutions design AI agents that honor their communities instead of replacing them. Same method, same discipline, same people-first approach.
GISTM.ai platform
AI for continuous tailings compliance, the industrial counterpart to Muiraquitãna.
AI agent for water balance
The same editorial discipline applied to water governance under MCA-WAF.
AI-assisted GISTM audit
Prioritized gaps and actionable recommendations, AI as a routing layer.
