The platform
One foundation. One workforce. Neither rests.
SparkJAR is the always-on operating platform — multi-agentic, multi-tenant, reactive. SparKLAW is the fleet of klaws working in the background: acquiring data, running campaigns, shipping code. The platform handles everything that comes in. The workforce handles everything it decides to go do.
Layer one
SparkJAR — the operating foundation
SparkJAR is the always-on backend. Not an assistant, not a chatbot — a running platform that handles live business logic. A customer opens the app: SparkJAR runs the workflow. A call comes in: SparkJAR routes the voice agent. A report is requested: SparkJAR generates it. It does not wait to be scheduled. It answers.
It is multi-agentic and multi-tenant. One API surface serves many businesses simultaneously, with every read and write scoped by client at the database row level. That isolation is not a setting you turn on — it is the architecture.
SparkJAR owns the reactive layer: the contextual memory, the named voice agents, the report and valuation engines, the coordinators that sequence document generation and signing. The glue that holds all of it together. Ata IQ guides sellers and buyers through the app. Sofia handles the phone. The HomeIQ engine produces valuations. The platform runs.
What SparkJAR handles
- • Voice agents — Ata IQ (in-app), Sofia (phone)
- • HomeIQ report and valuation engine
- • Document generation and e-signature coordination
- • Three-tier contextual memory
- • Multi-tenant API surface
- • Payment processing, mail, SMS
What it doesn't do on its own
- • Campaign-driven outbound work
- • Proactive data acquisition
- • Shipping new code and improvements
- • Running playbooks on its own schedule
That is SparKLAW's job.
Layer two
The connector — one governed chokepoint
Between SparkJAR and the business systems sits a governed adapter tier: roughly twenty lightweight services, one per domain — ERP, email, content publishing, CRM. Neither SparkJAR nor SparKLAW ever touches a business system directly. They call the connector. The connector holds the credentials, enforces multi-company isolation, and owns every read and write.
This matters in practice. You can run many clients on one platform without any client touching another's data. You can swap out an underlying business system and the agents keep working — they talk to the connector, not the system. And you have one place to audit every operation that touches real business data.
The connector is not glamorous. It is why the platform can be trusted at scale — one owner of credentials, one enforcement point, one place where isolation lives. It is also why onboarding a new tenant is configuration, not engineering.
Layer three
SparKLAW — the workforce you never see
Here is the hotel. The front desk answers guests. The bellman takes the bags. That is what guests experience. Behind them is the entire operation: housekeeping, kitchen, maintenance, facilities, the loading dock at 4am. You never see them. They are the reason the stay is worth coming back for.
SparKLAW is the part you never see.
A fleet of klaws — always in motion, always working, never waiting to be asked. Campaign-driven. Proactive. The kind of work a team of ten would get around to eventually, running continuously in the background while SparkJAR handles live traffic.
SparkJAR without SparKLAW
SparkJAR runs the business. Reactive. Multi-agentic. Very good at handling what comes in. But it doesn't campaign. It doesn't wake up on Tuesday and decide to acquire two years of county deed data because the foundation needs it. It doesn't identify a defect at 3am and open a PR before anyone noticed. It answers the phone. It does not make the calls.
SparKLAW makes it proactive. Not just reactive — campaigning. Not just running — improving. SparkJAR runs the business. SparKLAW improves it while it runs.
The klaw taxonomy
Conductor
The orchestrator. Listens, plans, confirms intent, then dispatches the right klaw for the job. Nothing in the fleet runs without the Conductor's direction. It is the only klaw that talks to a human before acting.
Creator klaws
Write and ship code. Clone a branch, run the build gates, open a PR, wait for a human to approve. SparkJAR improves itself because Creator Klaws identify, implement, and propose changes — then stop at the approval gate.
MDK — Meydomo Data Klaw
The data foundation for everything Meydomo does. MDK runs in the background, acquiring property records from primary government sources across all 67 Florida counties and keeping the dataset current. Without MDK, there is no valuation. Without valuation, there is no HomeIQ.
HomeIQ Outlook klaw
Buy-side analysis for properties outside the Meydomo system. When a buyer needs insight on any Florida listing, the Outlook Klaw does the analysis in the background — before the buyer asks the second question.
MOPS
Marketing operations klaw. SEO, content, outbound campaigns. Runs on its own schedule, driven by the playbook the Conductor set, executing the marketing work a team of three would handle on a good week.
Curator & ingestor
Data quality, continuously. The ingestor turns raw acquired records into clean, structured data the rest of the platform can reason over. The curator keeps that data correct over time. Good answers start here.
Contextual memory
What makes a set of agents into a platform is shared memory. Every klaw reads from and writes to the same contextual store. The voice agent on a Tuesday call knows what the operator klaw filed on Monday. The research klaw's findings are available to the whole system without a re-fetch.
Memory is layered by how fast it changes. That structure keeps a multi-agent system fast and correct at the same time.
Permanent
Identity, voice, and core context that never expires — who an agent is and what it knows about your business.
Warm
Mode-specific tools and patterns, held for the length of a session — the difference between an inbound call and an outbound campaign.
Volatile
The here-and-now: this call, this customer, the record updated thirty seconds ago. Refreshed constantly.
In a multi-tenant deployment, every read and write is scoped to a single client and enforced at the database row level. One platform. Many businesses. No client ever sees another's data. That isolation is architecture, not configuration.
The proof
Meydomo: the platform, deployed
Everything above is running in production at Meydomo, a licensed Florida real estate brokerage. SparkJAR runs every live transaction: voice intake, property valuation, document generation, e-signature coordination. SparKLAW runs the foundation underneath it: the property data, buy-side analysis, marketing campaigns. The connector governs every write to the business systems.
A licensed broker of record reviews and attests every transaction. The platform handles everything that doesn't require that license. Two founders stood up the brokerage in four weeks. First listing sold.
SparkJAR voice synth
Ata IQ
The in-app voice agent. Guides sellers through the 14-step listing workflow and helps buyers describe what they want — by voice, with full transaction context at every step.
SparkJAR voice synth
Sofia
The phone agent. Handles inbound and outbound calls for sales and transaction coordination, around the clock, with full memory of every prior interaction.
SparKLAW klaw
MDK — Meydomo Data Klaw
Acquires and maintains the property data foundation from primary government sources — parcel records, assessments, deed history across all 67 Florida counties. Runs in the background, always current.
SparKLAW klaw
HomeIQ Outlook
Buy-side analysis for every Florida listing — not just Meydomo listings. The Outlook Klaw analyzes the property before the buyer's second question.
Want this running inside your business?
We build and run SparkJAR + SparKLAW for one founder at a time, tuned to your vertical, your data, your operation.
Start a discovery call