# ZTABS — AI-first software agency in Houston building custom AI agents, RAG systems and full-stack products.

> Source: The Agents Index — https://theagentsindex.com/ztabs-llc (structured, researched, re-verified)
> Facts last verified: 2026-09-25

ZTABS is a 25-person AI-first software development company founded in 2015 that designs, builds and operates custom AI agents, RAG systems, SaaS platforms and mobile apps for clients. It also ships its own products, including the Chatsy support agent and EMRG sales agents.

| Fact | Value |
| --- | --- |
| Website | https://ztabs.co |
| Price | from $10,000/project |
| Free tier | Partly |

No free delivery work. A free scoping consultation precedes every engagement, and the site hosts free calculators and developer tools (ROI calculator, AI readiness assessment, JWT decoder, QR generator) provided 'as is'.

## What it does

**Writes code** · Yes — Custom development is the core engagement: they write the application as well as the agent, and say AI assists their own delivery from code generation through deployment.

> Founded in 2015, we have served 300+ clients, shipped 23 production SaaS products of our own (10 AI-powered), and use AI across our entire development process, from code generation to deployment.
> — https://ztabs.co/ · Homepage, 'What Is ZTABS?' section, first paragraph

**Debugs and fixes** · Partly — They diagnose defects as a QA and testing engagement, delivering reproducible bug reports and triaging failures; fixing is part of development or ongoing maintenance work rather than a standalone guarantee.

> We provide detailed bug reports with reproduction steps, screenshots, and severity classifications.
> — https://ztabs.co/services/qa-testing · 'How We Approach QA & Testing', third paragraph

**Writes tests** · Yes — Automated unit, integration, end-to-end, load and mobile tests are a named deliverable, written into the client's own CI pipeline so pull requests are validated automatically.

> We write tests that run in your CI pipeline (GitHub Actions, GitLab CI, CircleCI) so every pull request is validated automatically.
> — https://ztabs.co/services/qa-testing · 'How We Approach QA & Testing', second paragraph

**Reviews code** · Partly — Code review appears as automation they build for clients using AutoGen multi-agent review crews, not as a review service you buy for an existing diff; the review capability is the delivered system.

> Our team has delivered hundreds of code review automation projects with AutoGen, and the results speak for themselves.
> — https://ztabs.co/technologies/autogen-for-code-review-automation · 'Why AutoGen for Code Review Automation' section

**Whole-codebase work** · Yes — Repository-wide migrations are an explicit offering: they decompose legacy PHP, .NET, Java and desktop systems and modernise them module by module over six to eighteen months.

> Over 6–18 months, we migrate module by module until the entire application is modernized.
> — https://ztabs.co/services/legacy-app-modernization · 'How We Approach Legacy App Modernization', fourth paragraph

**Ships it** · Yes — Delivery runs past the diff into production: they build CI/CD pipelines that deploy to staging and production, and deploy agents into the client environment with monitoring.

> The first deliverable is usually a CI/CD pipeline on GitHub Actions or GitLab CI that runs linting, type-checking, unit tests, and integration tests on every pull request, then deploys automatically to staging and production with one-click approval.
> — https://ztabs.co/services/devops-services · 'How We Approach DevOps Services', second paragraph

**Done-for-you service** · Yes — This is a services business, not software you licence: a 25-person team designs, builds and continues to optimise the system for you after launch.

> We design and deploy autonomous AI agents that handle complex workflows, make decisions, and execute tasks across your business — from customer support and sales to operations and data processing.
> — https://ztabs.co/services/ai-agent-development · AI Agent Development page, hero paragraph

## How much it does unattended

**Runs autonomously** · Yes — The agents they build are described as completing multi-step workflows end to end, with human-in-the-loop escalation reserved for edge cases rather than every step.

> A well-built agent can qualify a sales lead by researching the company, checking CRM history, drafting a personalized outreach email, and scheduling a meeting — all without human intervention.
> — https://ztabs.co/services/ai-agent-development · 'How We Approach AI Agent Development', first paragraph

**Multi-agent** · Yes — Multi-agent orchestration is a named capability, with specialised research, analysis and action agents built on LangChain, CrewAI and custom orchestration layers.

> We design multi-agent architectures where specialized agents collaborate: a research agent gathers data, an analysis agent interprets it, and an action agent executes the decision.
> — https://ztabs.co/services/ai-agent-development · 'How We Approach AI Agent Development', second paragraph

**Agent permissions** · Partly — Guardrails are standard but their strictness is designed per engagement: rate limiting, content filtering, action confirmation and human-in-the-loop escalation are priced as scope, not a fixed policy model you configure.

> Every agent we ship includes guardrails — rate limiting, content filtering, human-in-the-loop escalation, and observability via tools like LangSmith and Helicone.
> — https://ztabs.co/services/ai-agent-development · 'How We Approach AI Agent Development', second paragraph

## Which models it runs on

**Claude** · Yes — Anthropic Claude is named as one of the foundation models they build agents on, alongside OpenAI and Google, and is listed in the EMRG product stack.

> ZTABS delivers production-grade AI solutions using GPT-4o, Claude, Gemini, and open-source models, with 10 AI-powered products shipped and 300+ clients served
> — https://ztabs.co/ai · 'What Are AI Development Services?' section

**GPT** · Yes — OpenAI models are the most frequently named across the site, with a dedicated GPT integration service and GPT-4o cited for complex reasoning in their routing approach.

> Token-level budgeting, routing cheap models for simple queries and GPT-4o / Claude for complex reasoning, plus response caching.
> — https://ztabs.co/ai · 'Our Approach to AI Development', point 3 'Cost control with model routing'

**Open models** · Yes — Open-weight models are supported both as integrations and as fine-tuning work, with Ollama listed for running Llama, Mistral and Phi on client infrastructure.

> Run open-source LLMs locally — Llama, Mistral, Phi, and more on your own infrastructure.
> — https://ztabs.co/ai · 'AI Technologies' section, Ollama card

**Model choice** · Yes — Model selection is presented as an engineering decision made against the client's accuracy, latency and cost constraints rather than a fixed vendor allegiance, including multi-model routing.

> We start with your business problem and select the right mix — foundational models, fine-tuning, RAG, or custom ML — based on accuracy, latency, and cost constraints.
> — https://ztabs.co/ai · 'Our Approach to AI Development', point 1 'Use case first, model second'

## Where you use it

**In your pipeline** · Yes — Their engineers work inside your pipeline: automated suites are wired into GitHub Actions, GitLab CI or CircleCI so every pull request is gated, and CI/CD construction is a separate DevOps offering.

> Tests integrated into your pipeline for continuous feedback on every change.
> — https://ztabs.co/services/qa-testing · 'What Our QA & Testing Includes', CI/CD Integration card

**In a browser** · Partly — There is no hosted workspace for the service itself; the only self-serve web surface is a set of free browser tools and calculators published on ztabs.co, offered without professional guarantees.

> We offer free tools on our website (such as the Marketing Plan Generator and SEO Checker). These tools are provided "as is" for informational purposes only.
> — https://ztabs.co/terms · Terms of Service, section 4 'Free Tools'

## Whose machine it runs on

**Self-hosted** · Yes — Systems they build can be deployed into the client's own environment, including on-premise and VPC-hosted options aimed at regulated healthcare and finance buyers.

> Data residency controls, PII redaction in prompts, audit logging, role-based access, and on-premise or VPC-hosted deployments for healthcare, finance, and other regulated industries.
> — https://ztabs.co/ai · 'Our Approach to AI Development', point 4 'Security and compliance by design'

**Open source** · Partly — Only one of their own products, the Beveled screenshot tool, is published as open source; the agency practice and its other products are not offered as source-available software.

> Open Source GitHub
> — https://ztabs.co/products · Products page, Beveled card statistics row

## What it costs to run

**How it meters** · Yes — Billing is per project or per developer-hour, not per seat or token: staff augmentation is quoted hourly by seniority while builds are quoted as fixed project ranges after scoping.

> Rates range from $35-$85/hour depending on seniority and specialization, with no hidden fees, no long-term contracts, and no recruitment charges.
> — https://ztabs.co/services/staff-augmentation · 'How We Approach Staff Augmentation', third paragraph

**Free tier** · Partly — You can start without paying but not use anything without paying: scoping consultations and the on-site calculators are free, while any delivery work is a paid engagement.

> Every engagement begins with a free consultation to scope your requirements and provide a detailed estimate.
> — https://ztabs.co/services/ai-agent-development · 'Ready to Get Started with AI Agent Development?' closing section

**Bring your own key** · Partly — Deployments can run in your own cloud or on-premise and integrate whichever model providers you hold accounts with, but the site never states explicitly that clients supply their own API keys.

> LLM orchestration with LangChain and CrewAI, vector search with Pinecone, Weaviate, and pgvector, and integration with OpenAI, Anthropic, and open-source models.
> — https://ztabs.co/ai · 'Our Approach to AI Development', point 2 'Production-grade orchestration'

## Buying it for a team

**A company can buy it** · Yes — The entire business is sold to organisations (startups, SaaS companies and enterprises), with engagements scoped by contract rather than individual self-serve signup.

> We work with startups building AI-powered MVPs, SaaS companies adding intelligent features to existing products, and enterprises modernizing legacy systems with AI agents and automation.
> — https://ztabs.co/ · Homepage, 'What Is ZTABS?' section, second paragraph

**Seat model** · Partly — Only the staff augmentation line is priced per person, scaling from one to twenty-plus developers month to month; no minimum seat count or ceiling is published for project engagements.

> Engagements are flexible — scale from 1 to 20+ developers, adjust monthly, and cancel anytime with 2 weeks notice.
> — https://ztabs.co/services/staff-augmentation · 'How We Approach Staff Augmentation', second paragraph

**Admin controls** · Partly — Constraints are engineered into each delivered system (role-based access, action confirmation, rate limits) rather than exposed as an admin console the buyer configures themselves.

> Input/output validation, content filtering, action confirmation, and rate limiting add $10K–$20K. Critical for production agents that take real-world actions.
> — https://ztabs.co/cost/ai-agent-development · 'What drives the price', Safety & Guardrails card

**Audit log** · Yes — Audit logging is standard in what they build, named both in the AI security approach and as an enterprise deliverable alongside encryption and role-based access control.

> Security is built in from day one: role-based access control, audit logging, data encryption at rest and in transit, and compliance with SOC 2, HIPAA, or GDPR depending on your industry.
> — https://ztabs.co/services/enterprise-software · 'How We Approach Enterprise Software', second paragraph

**Single sign-on** · Yes — SSO and SCIM provisioning are described as wired into enterprise builds from day one, with SSO integration listed as an enterprise security deliverable.

> 60+ enterprise-grade systems shipped across HRIS, billing, claims, and asset management — average deployment serves 500–10,000 named users with SOC 2-aligned controls and SSO/SCIM wired in from day one.
> — https://ztabs.co/services/enterprise-software · Enterprise Software page, proof statement under the hero

_Not established: Pooled budget_

## What happens to your code

**Data residency** · Partly — Residency is handled per build through on-premise or VPC deployment rather than published as a jurisdiction the vendor guarantees; the company itself operates from Houston, Texas.

> Data residency controls, PII redaction in prompts, audit logging, role-based access, and on-premise or VPC-hosted deployments for healthcare, finance, and other regulated industries.
> — https://ztabs.co/ai · 'Our Approach to AI Development', point 4 'Security and compliance by design'

**Certifications** · Partly — They implement SOC 2, HIPAA and GDPR controls in client systems and describe architecture as SOC 2-aligned or SOC 2-ready, but publish no certification held by ZTABS itself.

> Guardrails, human-in-the-loop, observability, and SOC 2-ready architecture from day one.
> — https://ztabs.co/services/ai-agent-development · 'Why Choose ZTABS for AI Agent Development?', Enterprise-Grade Reliability card

_Not established: Opt out of training, Getting out_

## Provenance

Every fact above comes from the vendor. Nothing here is independently corroborated yet.

## Coverage elsewhere

None recorded. Every fact above comes from the vendor and is not independently corroborated yet.

## What changed

- 2026-09-25 — Capabilities · Commercial terms · Free tier · Price

## Correcting this record

Anything marked "Not established" is a question we have not answered — it is not a No. If you represent ZTABS, or anything here is wrong or out of date, you can supply it with a source: the operations this site accepts from an agent are listed at https://theagentsindex.com/.well-known/agents.json — that catalogue needs no key to read, and each entry states its own auth.

Supplying evidence is free. Nothing that can be bought here changes what a field says, whether a gap is closed, or how this record is scored.