Enterprise Agents Hackathon: Powered by IBM Watsonx Orchestrate AI Labs, Tavily, and Redis

solosquard

Automate the detection of signup friction using AI agents that navigate like real users and report issues instantly.

RedisTavilyIBM

P

ContractIQ

We built ContractIQ, a 6-agent procurement workflow that turns contract documents into a decision-ready renewal package. IBM watsonx Orchestrate manages the full agent flow, including routing, parallel execution, confidence checks, and approval gates. Redis powers workflow state, inter-agent messaging, audit history, and cached memory across the pipeline. Tavily brings in live external research like vendor pricing, market signals, and competitive alternatives to ground the final recommendation.

RedisTavilyIBMPython, React, Typescript, AnthropiAPI, FASTAPI

Midas

OpenClaw Finance Agent — Enterprise AI that learns your business without seeing your secrets. Finance teams want AI that understands their vendor patterns, approval workflows, and GL structures — but compliance (GDPR, SOX, CCPA) blocks them from exposing sensitive data to train it. The result is that companies only use AI for suggestions. It's too risky to let it actually execute anything. We built a local-first AI system that solves this with comprehensive tokenization. Before any AI model sees the data, every sensitive field is replaced with a structural token. Microsoft Corp becomes VENDOR_001[TIER_LARGE]. $150,000 becomes AMT_LARGE[CFO]. Employee emails become APPROVER_LEVEL_CFO. PII is never extracted in the first place — it doesn't enter the pipeline at all. The model learns your company's patterns from tokens alone. The system uses two-tier AI routing. A locally fine-tuned model (LoRA on Mistral 7B, 4-bit quantized) handles 80% of routine tasks like invoice processing, spend analysis, and GL posting — under 500ms at $0.01 per task. Complex reasoning like forecasting and risk analysis routes to Claude API at $0.50 per task. Smart routing decides based on task complexity and model confidence. Every AI decision passes through MCP governance servers before execution. Is the approver authorized for this amount? Is the GL account valid? Is there budget available? Are there any policy violations? All gates must pass before the system acts. Every decision is logged to an immutable, hash-chained audit trail stored in tokens — not real data. SOX compliant by design. The architecture runs two isolated AI agents: Hermes handles coding and infrastructure (Docker, GitHub, codebase intelligence, Doppler secrets), while Midas handles financial analysis and compliance (governance, audit, GL validation, Tavily research, Redis). Each agent has its own Claude Code CLI wrapper and dedicated MCP servers. Namespace isolation ensures Hermes never touches financial governance and Midas never touches infrastructure. Training improves continuously through three channels: Tavily batch-researches public financial data, the /dream command generates synthetic training examples on demand, and Honcho from Plastic Labs automatically optimizes prompts based on real usage patterns. The system gets smarter with every interaction. Custom AI trained on your tokenized data. Runs on a laptop. Zero cost per local query. Full compliance audit trail. Policy-governed execution. Built in 48 hours for the IBM Hackathon, Procurement and Finance Track. https://www.loom.com/share/d05ac73cb5f94fb1b2ad99df0e1d32ea

RedisTavilyOpenclaw, Gemmma 4, Custom MCPs, Mac mini (Running Midas Local AI)

Vasco

Vasco is a mobile-first AI agent that gives building engineers the diagnostic depth of a senior chief engineer — without the $300,000+ salary. In this demo, a tenant in apartment 3B reports a heat pump faulting on high pressure. In under five minutes, Vasco identifies the unit, recognizes it as the third short-cycling event on the same riser this quarter, traces the supply chain upstream to a secondary condenser water pump issue, produces a sourced diagnostic walkthrough, and dispatches the contractor with the full diagnosis attached — all from the engineer's phone. Built on IBM watsonx Orchestrate, Claude Haiku 4.5 (pre-router), Redis (institutional memory), Tavily (live manufacturer documentation), Deepgram Nova-3 (STT), and ElevenLabs (TTS). Live app: https://vasco.one Youtube Video: https://youtube.com/shorts/HytYA_oQOZM?feature=share Github Video: https://github.com/hawkeye047/vasco/blob/main/vasco_video.mp4

RedisTavilyIBMIBM watsonx Orchestrate (agent runtime) Mistral (composer model inside watsonx) Claude Haiku 4.5 (pre-router) Redis (memory) Tavily (live docs retrieval) Deepgram Nova-3 (STT) ElevenLabs (TTS) Next.js (frontend) Vercel (hosting — inferred from .one deploy, confirm)

willow

TrustOps Sentinel AI is a real-time AI governance platform that intercepts and evaluates agent actions before execution. Using a policy enforcement layer, contextual intelligence, and blockchain-backed auditing via Algorand, it ensures proactive prevention, explainability, and secure compliance for enterprise AI systems. Demo: https://drive.google.com/file/d/17iV6LugpKBKiDnN5EqTe2ceei4dCYTjW/view?usp=drivesdk

RedisTavilyIBMMySQL, AlgoRAND

H

nocept

MAIN DEMO: https://drive.google.com/file/d/1AWSg1w5srtz2kc2d1u2acTc1YYBGPmcT/view?usp=sharing STREAMLIT UI DEMO:https://drive.google.com/file/d/1oVChH6PqPA2oJ0zWeaNZ5tTYfMuagV1G/view?usp=drive_link DESCRIPTION: Enterprise Accounts Payable workflows generate significant volumes of invoice exceptions — discrepancies betweenpurchase orders, goods receipts, and supplier invoices that require manual investigation and resolution. These exceptions are difficult to resolve programmatically because the root cause typically exists outside the ERP system, in emails, verbal agreements, or undocumented supplier changes. NOCEPT is an agentic system built on IBM watsonx Orchestrate that implements a six-step sequential tool-call pipeline for autonomous exception resolution. Orchestrate manages the agent loop, invoking tools in order: exception classification via three-way match analysis, variance tolerance gating, Redis-backed historical similarity scoring against past approved cases, keyword confidence scoring across linked emails and transcripts, and Tavily-powered external web research for corroboration. Each tool writes its decision to Redis with a TTL, and the resolution tool reads whichever gate fired first. Short-circuit logic terminates the pipeline early on auto-approval. If no gate reaches threshold, the agent escalates with all intermediate evidence preserved in the audit log.

RedisTavilyIBM

sebastian roda

DEMO VIDEO LINK: https://youtu.be/RzUFxBRLbJo OrgSim is a decision rehearsal engine for internal business choices. Instead of automating tasks, it asks: “If we make this decision now, how will different departments react over time, and where will conflicts emerge?” This is novel because there has not been a orchestration system that also incorporates a predictive component into it. utilizing a swarm simulation of agents from different departments to speak to one another allows for this simulation of scenario to identify conflicts and to avoid them. Most tools tell you what happened or automate what you already do. OrgSim tries to predict organizational ripple effects before a decision is made.

RedisTavily

D

Agents of Agents

In today’s music industry, the challenge is no longer access to talent, but the ability to identify the right talent early enough to act on it with conviction. As the volume of independent releases, social signals, and fragmented audience data continues to grow, A&R teams are forced to sort through an overwhelming amount of noise to find the few artists who represent real commercial opportunity. We believed there was room for a system built not for casual discovery, but for strategic decision-making at the enterprise level — which is why we built Sonar; an elite A&R intelligence agent designed for record labels and music companies such as Warner, Sony, and Universal, as well as distributors and talent-focused platforms such as DistroKid, to help their teams scout emerging artists with greater speed, confidence, and precision. Rather than replacing the instincts of an experienced A&R executive, Sonar is built to strengthen them. The platform functions as an intelligence layer that helps companies surface early-stage artists with meaningful momentum before they become obvious to the rest of the market. By leveraging advanced metrics, cross-platform validation, and predictive analytics, Sonar allows decision-makers to evaluate not just who is visible in the moment, but who is demonstrating the kinds of patterns that signal future breakout potential. In doing so, it enables labels and talent organizations to make smarter, earlier, and more defensible investments — reducing guesswork while increasing the likelihood of identifying artists whose growth can translate into long-term value for the business. It didn't let me add a video for some reason. Powerpoint slides https://1drv.ms/p/c/9c5cb125c160f90c/IQBkYlTV340KQo_SfvMk7w14AbYKUPR5qNnbM8ut_9n--OA?e=RcI86n

RedisTavily

Diamond Team!

AI agent that resolves invoice exceptions for accounts payable teams. Upload an invoice — image or text — and the agent extracts structured data, matches it against purchase orders, searches the web for vendor verification, and applies your company's rules to deliver a decision (approve, resolve, or escalate) with full reasoning and audit trail. Configurable exception rules, an LLM-as-judge for custom policies, and a human review queue close the loop between AI recommendations and human oversight. Live demo: https://genuine-dream-production-1287.up.railway.app Deck: https://genuine-dream-production-1287.up.railway.app/deck You can try it out on the live site and also download a very short demo video here: https://genuine-dream-production-1287.up.railway.app/demo.webm

RedisTavilyOpenRouter

Pivt

Pivt is an agentic supply chain crisis management platform built on IBM watsonx Orchestrate, designed to help logistics companies detect, respond to, and resolve delivery disruptions in real time, starting with weather-driven events. When a live NWS weather alert intersects an active delivery corridor, Pivt's three-agent pipeline activates automatically: the Optimizing Agent calculates reroute options with full cost and SLA analysis, the Facility Agent identifies the nearest viable drop facility if rerouting isn't justified, and the Driver Agent notifies the driver and pushes an updated route directly to Google Maps, all within 15 minutes. The platform uses Tavily for deep, real-time research on corridor conditions, weather headlines, and logistics disruption signals, giving every agent accurate and current situational context before making a recommendation. Redis serves as the platform's memory layer, maintaining live state across the agent pipeline so each agent picks up exactly where the last one left off, preserving shipment context, dispatch decisions, and route history across the full response flow without losing continuity. Together, the three agents act as a supply chain monitoring and response system and a built-in CRM for crisis resolution, turning what is today a 4–8 hour manual scramble into a fully orchestrated, auditable, 15-minute response. You can find the presentation link here: https://drive.google.com/file/d/1wPigTtPBGt5ioSkyl7xfD3oa7jMLKU-6/view?usp=sharing https://hel1.your-objectstorage.com/hackersquadcontent/project-recordings/project_rec_cmnw32fvc00iyr00kk6fq5ana-2026-04-12T184912.mp4

RedisTavilyIBM

Toni

CyberSearch is an agentic security analysis system that evaluates every code commit using real-time external intelligence, contextual LLM reasoning, and persistent repository memory. Unlike traditional scanners that rely on static CVE databases, CyberSearch integrates live threat intelligence via Tavily, processes full commit context using IBM WatsonX, and maintains historical state with Redis to enable continuous learning across analyses. The system runs as an event-driven pipeline triggered by GitHub webhooks, performing multi-step reasoning to determine whether newly introduced changes are actually exposed to emerging vulnerabilities. Each analysis produces a risk score, classification, and actionable decision (allow, review, or block), along with a detailed explanation. By combining real-time data, agentic reasoning, and stateful memory, CyberSearch reduces detection latency from days to seconds and enables proactive security decisions directly inside the development workflow. The goal is to move application security from reactive scanning to continuous, intelligent, and context-aware defense. https://drive.google.com/file/d/1w7SfqttmU4J3LKmqV15LJOjURxP8-XgL/view?usp=sharing

RedisTavilyIBM

V

DataForce

ProcureIQ is an autonomous AI agent that prepares software contract renewals end-to-end in under 20 seconds — detecting upcoming renewals, pulling negotiation history from memory, gathering live market intelligence, computing BATNA (Best Alternative to a Negotiated Agreement), generating a cited negotiation brief, drafting a vendor outreach email, and routing for internal approval — all without human intervention. Pitched as: "The autonomous procurement analyst for mid-market companies that can't afford Tropic ($36K/yr). Self-hosted for ~$600/yr." 10-step ReAct loop powered by IBM Watsonx Orchestrate ADK (tool registration), IBM Watsonx.ai Granite-3-8b-instruct (brief + email generation), IBM Watsonx slate-30m embeddings (semantic cache), Tavily (4 parallel vendor intelligence queries), and Redis (SemanticCache for 7-day vendor intel + long-term negotiation memory via Agent Memory Server). Key capabilities: - Live market intelligence: pulls competitor pricing, vendor red flags, and financial signals from 20+ sources per vendor via Tavily - Semantic cache: RedisVL SemanticCache stores intel with IBM Watsonx embeddings — cache hits return in <200ms, saving Tavily credits - Long-term memory: Redis Agent Memory Server retains negotiation history across sessions; agent cites past outcomes when computing leverage - BATNA calculator: computes target price, walk-away price, discount %, leverage points, and a 0–100 confidence score from contract + usage + intel data - Fully template-fallback: demo runs without LLM keys; IBM Watsonx unlocks LLM-generated briefs and emails Demo scenarios: A) Datadog (strong leverage) — 48 unused seats + competitor pricing → 20%+ target discount B) Snowflake (weak leverage) — 95% utilization → agent honestly says "weak position" C) Cache hit — re-run any vendor → Redis returns intel in <200ms, no Tavily credits used

RedisTavilyIBMFastAPI,Streamlit,IBM Watsonx Orchestrate ADK, ibm-watsonx-ai SDK, Python 3.12, uv

ARF Squad

An enterprise access control agent built on IBM watsonx Orchestrate that uses the Agentic Reliability Framework (ARF) for Bayesian risk scoring. It evaluates access requests and returns approve/deny/escalate decisions with a full audit trail (risk_score, justification, timestamp). The agent integrates with Redis for audit persistence and Tavily for policy retrieval (optional). Demo shows three scenarios: admin → approve (risk 0.15), intern → deny (risk 0.85), contractor → escalate (risk 0.55).

IBMAgentic Reliability Framework (ARF), OpenAPI, Python, JSON

AIGatekeeper

Shadow AI Gatekeeper tackles the shadow AI problem — employees adopting unapproved AI tools that expose the company to security and compliance risks. Instead of blanket bans or slow manual reviews, we built three agents on IBM watsonx Orchestrate that automate the entire vendor approval lifecycle. The Gatekeeper Agent evaluates vendors in real time by searching the web via Tavily, extracting structured evidence with LLM, and scoring against a deterministic company policy — returning instant green/yellow/red decisions. The Comparison Agent lets employees compare multiple vendors side-by-side. The Monitoring Agent continuously scans approved vendors for security incidents and flags risk changes. Yellow-path decisions escalate to human approvers via a Next.js dashboard with full evidence visibility, activity timelines, and confirmation dialogs. Every decision is explainable (extraction is ML, scoring is not), every action is auditable via Redis Streams, and repeated questions are answered instantly via RedisVL semantic cache. One Redis instance powers five jobs: case state, audit log, evidence cache, semantic cache, and dashboard queries.

RedisTavilyIBM

RAGnos Labs

Vendor Intake Copilot is a Watson-native enterprise agent that turns a messy vendor intake or vendor backlog into a first-pass decision brief, with clear reasoning, a weighted scorecard, supporting citations, and a clean downstream handoff. The workflow starts from either a single vendor request or a CSV backlog import. The system cleans and normalizes the intake, pulls live external evidence on each vendor, checks that against internal security and procurement requirements, and produces a structured CASE packet with decision status, risk tier, blockers, missing evidence, recommended controls, and next owner. IBM watsonx Orchestrate is the user-facing UI and orchestration layer. Redis provides the durable retrieval and review context layer for internal policy, prior review memory, and case continuity. Tavily provides live web research and extraction so the agent can ground vendor decisions in current public evidence rather than generic model guesses. For the demo, we show: CSV backlog preview and cleanup warnings, shortlist generation, queue summary, individual case review, citations, and a downstream Slack handoff packet for the next team. The point is to compress a slow cross-functional vendor review workflow into a faster, auditable first pass without removing the human approval gate.

RedisTavilyIBMNode.js, Python, Docker Compose, OpenAPI, Slack webhooks, CSV import/normalization pipeline, cloudflared tunnel (automatic fallback: ngrok)

SancScreen

SancTrust's agentic AI system automates counterparty financial-crime risk assessments by connecting to live lookup databases and orchestrating specialized research agents that investigate entities across sanctions lists, adverse media, enforcement actions, regulatory databases, and even complex cross-jurisdiction ownership structures. It turns that research into clear, structured due diligence reports with a risk rating and a traceable evidence trail, giving organizations a scalable way to run rigorous compliance screening without the overhead of manual research. A human-in-the-loop review layer keeps the process regulatorily defensible by ensuring findings are verified and backed by auditable evidence. Although it is purpose-built for financial-crime risk assessment, the underlying orchestration and architecture are flexible enough to support other complex due diligence workflows across industries. Slides: https://v0-presentation-generation-beta-six.vercel.app/ Demo Video youtube : https://youtu.be/JXubU0oLc3s googleDrive: https://drive.google.com/drive/folders/1tft11it1CLPwO381L0kn5RkwGaPReKVj?usp=sharing

RedisTavily

Elite

AI-assisted home search: type what you want in the browser, and an Express API runs Tavily Search + Extract to pull real pages and a short summary

Tavilyexpress.js

The New Guard

Context_Leash is the natural merger of two complementary ideas. One project imagined a personal context firewall that protects individual users from their own AI agents — automatically redacting secrets, blocking malicious exfiltration, and giving every person a cryptographic leash on their second brain. The other built a governed decision engine for enterprise access requests, turning vague agent actions into auditable approve, deny, or escalate outcomes. We combined them. Context_Leash now brings the discipline of a context firewall into enterprise workflows. It sanitizes sensitive data before an agent can see it, prevents unauthorized leaks, enforces policy-as-code rules, and delivers clear, risk-aware decisions with full audit trails. The result is an agent that remains powerfully useful yet can no longer betray its user — whether that user is a single developer or an entire organization. In the end, we gave agents what they needed most: freedom to act, held firmly by a trustworthy leash.

RedisTavilyIBM

404 Survey not found

We built a Voice of Customer Intelligence Agent that transforms traditional survey data pipelines into real-time, AI-driven decision systems. The agent continuously ingests customer feedback, analyzes it using IBM Watsonx Granite for sentiment, topics, urgency, and PII redaction, and enriches each signal with live industry context via Tavily. A supervisor agent evaluates each response, identifies critical signals such as churn risk or legal threats, and escalates them instantly to Slack, enabling human-in-the-loop intervention. Built on an event-driven architecture using Redis Streams, the system is designed for real-time processing, reliability, and scalability. By reducing crisis detection time from days to minutes and automating manual analysis, our solution bridges the gap between raw data pipelines and actionable business intelligence. https://www.youtube.com/watch?v=elC7RVB_ZJc

RedisTavilyIBMQualtrics mock, fastapi