Based in Lahore · UTC+5
5 systems shipped · Prices and method public
Fixed scope · Fixed price
Open book Production AI agents for property and field-service operations

Work orders. Tenant tickets. Dispatch.
Handled by agents.

Demo-ware dazzles in a meeting and never runs on Monday. We build agents that do: each one ships with a fail mode, a rollback path and the eval that proves it works. Our prices, method and evals are public, so you can check the work before the first call.

First Agent $8K — 15K · 4 — 6 weeks Production System $15K — 30K · 6 — 10 weeks
Example workflow

A maintenance request, handled.

Incoming tenant message

“The kitchen sink is leaking.”

  1. Understand the request

    Read the message and property context.

  2. Route the right work

    Check urgency and assign maintenance.

  3. Prepare the next step

    Create a ticket and draft a tenant update.

Eval gate before action.Your team reviews the exceptions.
Stack we ship on
PythonLangGraphpgvectorClaudeOpenAICloudflare WorkersD1Next.jsAWS
01Services

Two engagements. Both fixed-scope, fixed-price, no hourly billing.

First Agent
One workflow, live in production.
$8K — 15K
4 — 6 weeks

For the one workflow that eats your team's week: tenant tickets, quotes, service calls. One agent, or a small team of agents, running in production and measured.

  • A real agent on your workflow, not a chatbot wrapper
  • Fail mode, rollback path and the eval that proves it works
  • Production behind feature flags you co-own
  • Two iteration rounds inside the SOW
  • 30-day support after handoff
Scope a First Agent

Terms for both. SOW signed before any work. 50% at SOW signing (non-refundable), 50% at production handoff. IP transfers on final payment. Change requests get their own SOW. The 30-min scope call is free, and there is no paid discovery phase.

02Method

The Deploy Ladder. Map, build, wire, ship, with every rung agreed before any code.

First Agent4 — 6 weeks Production System6 — 10 weeks CadenceDaily written update · Friday demo
01 · Map

Map the agent surface

Workflows, decision points and data sources. The first two days are discovery, inside the fixed price. We agree the rungs to production before any code.

Architecture + agreed rungs
02 · Build

Build the spine

Orchestration, memory, retrieval and tool calls go in first. Stateless before stateful, deterministic before probabilistic. Every PR deploys to a preview URL.

Rung 1 · Internal preview
03 · Wire

Wire the agents

Domain agents wired into the spine, into your tools, and into the eval suite. Every agent ships with a fail mode, a rollback path and the eval that proves it works.

Rung 2 · Staging on main
04 · Ship

Ship to production

Live behind feature flags we own jointly with you. Handoff: runbook, eval report, architecture diagram, Loom walkthrough and a 30-day support calendar. Code is yours.

Rung 3 · Production behind flags

First Flip. The first feature flag flipped in production, usually within ten working days of kickoff. Every Friday, a demo: a production or staging URL, plus a Loom under five minutes.

Read the Deploy Ladder
03Work

Five systems shipped. Property operations and field service first.

Verticals, durations, stacks and architecture are real. Outcome numbers are published only with client sign-off, and none are signed off yet. Client names are withheld. How we publish proof.

Property management · Operations 8 weeks · Production System

Tenant-ops copilot for a US property management platform.

Multi-agent system handling lease ingest, maintenance triage and tenant comms. Three specialist agents wired through a routing orchestrator, with eval gating before any action.

Metric: Ticket resolution time
Number published when the client signs off.
Next.jsClaudepgvectorAWS
Intake Router Lease Maint Comms Eval Action Read case study
Field service · Dispatch10 weeks · Production System

Autonomous dispatch agent for a North American field-service network.

Routing, scheduling and customer-comms agents wired into the existing CRM.

Metric: After-hours coverage and added headcount
Number published when the client signs off.
FastAPIOpenAIPostgres
Read case study
Also shipped Construction · EdTech · Document AI
Construction · Estimation6 weeks · Production System

Quote-generation system for a regional contractor.

Doc-aware agents reading drawings, pulling supplier prices and drafting client quotes.

Metric: Quote turnaround time
Number published when the client signs off.
PythonClaudeMilvus
Read case study
EdTech · Learning5 weeks · First Agent

Exam-prep platform with PDF chat and MCQ generation.

Conversational document agent over student coursework, with auto-generated practice questions.

Metric: Time to MVP in production
Number published when the client signs off.
Next.jsOpenAISupabase
Read case study
Document AI · Arabic + English7 weeks · Production System

Arabic and English RAG over about 100 regulatory PDFs.

Open-weight Arabic LLM stack with hierarchical retrieval over Milvus.

Metric: Top-3 retrieval accuracy on the eval set
Number published when the client signs off.
MilvusLlamaPython
Read case study
04Build vs buy

When to hire us. When to use Lindy or Relevance instead.

No-code agent platforms are good. For some teams they are the right answer. We will tell you which one you are on the scope call. Here is the honest version.

Scroll across to compare all three options

Compare time to delivery, ownership, deployment, and fit.
No-code platforms Generic dev shop Codantrix Labs
Time to production Days 2 — 3 months 4 — 10 weeks, fixed
Custom logic above platform ceiling Limited Maybe Yes
Owned codebase No Yes Yes
Production observability + evals Basics Rare Default
On-prem or VPC deploy Hosted only Possible Yes
Best for Sales / support under platform ceiling Generic web + AI features Ops workflows that need custom logic, owned code and evals
05Studio

Since 2023, in Lahore. Registered as Codantrix Labs in 2025. Small on purpose.

Why the studio exists

In 2023, Hassan was building systems alone for an outsourcing firm that billed clients $2 — 3K for each one and paid him 30 — 50K PKR. The clients were paying mostly for the layers between them and the person doing the work. Codantrix Labs exists so the engineer who scopes your system is the one who ships it.

So we stay small on purpose. No layers between you and the engineer writing your code, and never more than three engagements running at once. SOW first, a written update every day, IP transfers on final payment.

Since 2023. Registered as Codantrix Labs with the SECP in 2025. Based in Lahore, working with teams in US and EU time zones: async-first, with one weekly sync. More on the studio.

Building since
2023
Systems shipped
5
Verticals shipped
5
Engagements at once
3 max
HM
Founder · Engineering lead

Hassan Ali Mehdi

Seven years coding, three on production agentic systems. Former president of the AI Society at The Superior University, Lahore. BSCS '26. Builds agentic systems end to end: architecture, agents, frontend, deploy. Treats the SOW like a contract, because it is one.

06Common questions

Questions before we build. Answered before the call.

Q.01
How is pricing structured?
Fixed scope, fixed price, no hourly billing. A First Agent is $8K — 15K; a Production System is $15K — 30K. The SOW is signed before any work begins. 50% at SOW signing (non-refundable), 50% at production handoff. Change requests get their own SOW.
Q.02
Do we pay for discovery?
No. The 30-min scope call is free and scopes the work. There is no paid discovery phase before signing. The first two days of Map are discovery, inside the fixed price.
Q.03
How soon does something run in production?
First Flip, the first feature flag flipped in production, usually lands within ten working days of kickoff. Usually, not always. Every Friday you get a demo: a production or staging URL plus a Loom under five minutes.
Q.04
Who owns the IP and code?
You do. IP transfers on final payment, and the code is yours. The handoff box carries the runbook, eval report, architecture diagram and a Loom walkthrough. There is no platform of ours you depend on after handoff.
Q.05
How do you work across time zones?
Async-first. A written update every day, plus one weekly sync. We are in Lahore (UTC+5) and work with teams in US and EU time zones.
Q.06
What if we should use Lindy or Relevance instead?
Sometimes that is the right answer. We will tell you on the scope call. If your workflow fits inside a no-code platform, hire us only when you need custom logic, owned code, or VPC deploy that platforms cannot give you.
Q.07
What happens after handoff?
30-day support after handoff is included, on a support calendar you get in the handoff box. After that, new work gets its own SOW, fixed scope and fixed price like the first.

You bring the
workflow. We
ship the agents.

A free 30-min scope call about your workflow. By the end you will know whether we are the right fit, what it will cost and how long it will take. Within 24 hours you get a one-page Build sketch.

Double opt-in: nothing arrives until you confirm. One-click unsubscribe in every email. Our privacy notice was updated on 23 September 2026 to cover Field notes.