Continuum

Inspiration

Traditional B2B procurement is crippled by manual latency. When a data center router fails or a viral brand runs out of stock, companies bleed revenue waiting for employees to manually find vendors, check availability, and draft purchase orders. We built Continuum to automate this entire pipeline, connecting real-time system alerts directly to instant B2B procurement.

What it does

Continuum is a zero-touch autonomous continuity engine that resolves infrastructure and inventory failures without human intervention: Monitors: Streams live telemetry metrics (e.g., stock = 0, server_status = 503) to immediately catch boundary failures. Discovers: Uses the open agentic web to find alternative vendors dynamically, natively resolving HTTP 402 payment walls using micro-payments. Procures: Synthesizes structurally compliant B2B Purchase Orders (POs) matching vendor specifications and submits them to fulfillment endpoints. Learns: Logs successful resolutions to a local memory matrix, bypassing cold discovery to fix subsequent identical errors instantly.

How we built it

We engineered a modular state-machine runtime with local flat-file storage: Nexla ADK: Live telemetry ingestion and alert stream filtering. Zero.xyz: Dynamic tool discovery and keyless x402/MPP micro-payments. Pomerium: Zero-trust identity proxy gating internal database mutations. Akash Network: High-performance decentralized compute hosting the core agent loop.

Challenges we ran into

Dealing with the chaotic, unstructured nature of enterprise data. Telemetry streams from legacy infrastructure databases are incredibly noisy, and every alternative B2B vendor represented inventory schemas, pricing tiers, and shipping SKUs in completely incompatible formats. Resolving paywalled API structures under strict time pressure while normalizing these messy, mismatched datasets on the fly without breaking our automation state machine was a major hurdle. We solved this by implementing an adaptive schema-mapping translation layer that sanitizes, standardizes, and validates noisy incoming telemetry before the agent constructs the final transaction parameters.

Accomplishments that we're proud of

Immediate Business Value: Solving a high-stakes, multi-million dollar business continuity problem. Continuum prevents catastrophic revenue leakage during system crashes and recaptures immediate sales margins during stockouts by dropping recovery latency from days to milliseconds. The Learning Layer: Successfully demonstrated a "Run 2" optimization scenario, dropping execution latency from a 15-second cold discovery chain down to an instant 2-second resolution on the second identical failure. True Autonomy: Achieving complete zero-touch execution during a live split-screen stage demo while the target application environment was actively sabotaged.

What we learned

Reliable automation requires completely separating non-deterministic LLM reasoning from deterministic safety and authentication gates. Utilizing a runtime registry like Zero.xyz proved that hardcoding thousands of fragile API hooks into enterprise software is obsolete; agents can securely discover and purchase the tools they need on demand.

What's next for Continuum

Cross-Tenant Intelligence: Building a zero-knowledge shared network for independent Continuum nodes to securely pool anonymized vendor reliability and speed metrics. Predictive Procurement: Ingesting global weather and maritime feeds to proactively purchase backup supply before a delay ever registers in warehouse inventory.

Built With

• agentic-workflows
• akash-network
• b2b-procurement
• claude-code
• mpp
• nexla-adk
• next.js
• pomerium
• python
• tailwind-css
• typescript
• x402
• zero.xyz