The front door to URBLD's AI ecosystem.

Why URBLD is built for the AI era

Most field service software was designed for humans clicking through screens. URBLD was designed from day one for a world where AI agents — some built by URBLD, some built by third parties, some built by the business owner — do a growing share of the work. That shift is not cosmetic. It changes how identity, permissions, data access, auditability, and error recovery must be engineered. The AI Readiness Hub is the single place where humans, developers, and autonomous agents can understand what URBLD exposes, how to use it safely, and where the guardrails live.

Every capability you can click in the URBLD interface is also available to an authenticated agent through the same governed API. A human dispatcher rescheduling a job, Jarvis rebalancing a route after a cancellation, and an external agent reacting to a webhook all take the same path through the platform — the same authorization checks, the same tenant isolation, the same audit trail. There are no shadow endpoints for AI and no privileged bypass for internal automations.

The four layers of URBLD's AI stack

URBLD's AI surface is intentionally layered so you can plug in at the level that fits your team's technical depth.

• Built-in AI products. Jarvis, AI Receptionist, AI Follow-Up, and the AI Photo Estimator are turnkey — install URBLD, connect your phone number, and they start working on your pipeline immediately. No prompt engineering required.
• Agent infrastructure. Identity, scoped OAuth tokens, per-organization rate limits, and a tenant-visible audit log. This is the layer that makes it safe to hand a bot the keys to a business's operational data.
• Developer API. A REST surface covering the full lead-to-payment lifecycle, described by OpenAPI 3.1 and consumable by any HTTP client, agent framework, or workflow tool.
• Machine-readable discovery. /openapi.json, /.well-known/ai-capabilities.json, /agent.json, and /llms.txt let modern agent frameworks auto-generate tools without a human writing glue code.

Grounding: no hallucinated metrics, ever

URBLD's internal manager agent, Jarvis, follows a strict quantitative grounding rule: every count, list, and metric must come from a structured tool call against the tenant's own database. Jarvis cannot make up how many leads are open, how much revenue is outstanding, or which technician is closest to a callback. Third-party agents building on URBLD are strongly encouraged to follow the same discipline — read from the API, cite the endpoint, and refuse to summarize numbers from prose. The AI Agents Guide documents the pattern in detail.

Safety: agents inherit permissions, not bypass them

Every agent-issued mutation flows through the same row-level security policies, role checks, and organization boundaries that protect human users. An agent authenticated with leads.read cannot read a job. An agent scoped to one organization cannot see another organization's data. Destructive actions — deletions, contract signings, payment captures — write a signed entry to the tenant's audit log with the resolved agent identity, the scope stack that authorized the call, and the before/after state. Owners can trace exactly what any agent did, when, and why, and revoke a client with one click.

Discovery-first: agents find URBLD without your help

Modern agent frameworks — from Anthropic's Model Context Protocol to Google's Agentic Resource Discovery — expect to be able to point at a domain and auto-configure. URBLD ships every discovery document those frameworks look for. A customer's ChatGPT, Claude, or in-house LangChain agent can be pointed at urbld.com and immediately understand the capability surface, the authentication model, the rate limits, and the risk tiers of each action, without a human writing a single line of integration code.

Where to start

If you run a service business and want to use URBLD's AI, start with the AI Receptionist or Jarvis feature pages. If you build integrations or agents, start with the Developer Portal and the AI Agents Guide. If you are an AI agent reading this page directly, fetch /.well-known/ai-capabilities.json and proceed from there.