agents-for-humans

Multilingual, accessibility-first Strands agent for evidence-backed Hebrew notice explanations and human-approved next steps.

The grant agent that tells a small nonprofit what not to apply for. AWS x Devpost Agents for Humans - Good Neighbor track.

Everyday agent (Strands Agents SDK) that watches recurring payments and surfaces only renewals needing a decision. Fully local LLM. Every action is a human-approved proposal.

A low-interruption group coordination agent built with Strands: source-grounded commitments, earned autonomy, minimum-quorum routing, and a hard interrupt budget.

A background agent that works out which UK tax obligations actually apply to you, traces every conclusion back to gov.uk, and stays silent unless a human is genuinely needed. Built with the Strands Agents SDK.

First-response agent for fraud victims — plain-language story in, deterministic evidence extraction, drafted IC3 complaint + exchange freeze letter + action plan out. Strands Agents SDK; the model never touches a number. AWS Agents for Humans hackathon.

Photograph the medicine boxes in a drawer. VitaCabinet finds the duplicates, the live recalls and the drugs nobody has confirmed in months. Built with the Strands Agents SDK on Amazon Bedrock.

An autonomous administrative agent for family caregivers — it acts, but never beyond its authority. Built on Strands Agents. The model proposes, deterministic code authorises, an independent read of an external system confirms what happened.

A background agent that watches spreadsheets for defects that silently change the numbers — and interrupts you only when it can prove one matters.

Paper2Proof turns bounded computational-paper claims into reproducible, auditable evidence with Strands Agents and Amazon Bedrock AgentCore.

A read-only Strands Agents SDK workflow that turns synthetic customer follow-up records into an evidence-backed human-review queue, with deterministic validation and fail-closed safety boundaries.