Mention @MARi
Ask a question in an approved Signal group.
Metispace application / Signal & local AI
MARi
The answer is already
in your documents.
MARi is a private knowledge assistant for Signal that runs document search, embeddings, and language-model processing locally on your device. Teams receive clear, source-backed answers from approved internal documents without sending that material to a cloud AI model.
See how MARi works ↓Operations Team12 members
Today
Source: Evacuation Policy · section 4.2
Local chat, embedding, and vision models keep document indexing and AI processing on the MARi device, with tools to manage sources, test responses, monitor activity, and review answers.
How MARi works
Ask once. MARi does the finding.
No new portal. No new workflow. Your team asks in Signal and MARi returns a grounded answer with its source.
Find the right source
MARi searches the active local document collection and selects the most relevant passages.
Answer with evidence
A local AI model writes a clear reply and MARi posts it back with source references.
Signal group
Local document index
Local AI
Source-backed reply
TM@MARi what does the policy say about refunds?
MRefund requests must be submitted within 30 days.Source: Refund Policy · section 2.1
Knowledge that stays useful.
From policy PDFs to photographed field guides, MARi turns the material your team already trusts into answers they can use.
Works where teams already talk
Mention @MARi in an approved Signal group. No separate app or new account for your team.
Grounded in your sources
Every answer is built from the active document collection and returns with source references.
Understands real-world documents
Index PDFs, Word files, notes, images, and scans. Local vision processing can transcribe photographed pages.
Reviewable by humans
See the full reply, retrieved sources, response time, audit notes, and human feedback in the local admin view.
Collections keep different knowledge domains separate—operations, HR, travel, finance, or the work unique to you.
Recent replies
All collections⌄All sources⌄
QuestionSource matchAuditFeedback
What’s the per diem rate for rural travel?Travel Policy.pdf
p. 142.4s✓ Correct
p. 142.4s✓ Correct
Do we require a safety brief before field work?Field Guide.docx
p. 83.1s× Wrong source
p. 83.1s× Wrong source
How should incident reports be submitted?Reporting Policy.pdf
p. 52.7s○ Needs review
p. 52.7s○ Needs review
Designed for privacy
Private by design.
Clear about the boundary.
MARi combines Signal’s end-to-end encrypted messaging with local document search and local AI processing.
Protected in Signal
Signal conversations are always end-to-end encrypted between participants. MARi joins as a group participant, so it can read and answer messages in groups where it has been added.
Processed locally
MARi uses local Ollama language, embedding, and vision models. Documents, vector indexes, retrieved context, OCR transcripts, model calls, and review logs remain on the MARi device; no cloud AI model is required.
Controlled by your team
Approved group access, device security, document permissions, and human review remain part of the operating model.
Encryption protects messages in transit. The security of the MARi device and who belongs to each Signal group still matters.
Important boundary
Signal is a third-party service. MARi is not affiliated with or endorsed by Signal.
MARi provides source-backed assistance from the document collection configured by its operator. It does not replace human judgment, policy ownership, access control, or review of high-impact decisions.