The facts, with their sources

What it does

Writes code: PartlyITECS delivers custom software through an AI-assisted build that a senior architect supervises, so code is authored by the agency rather than by a tool the buyer drives.Custom software delivered through the ITECS AI-DevOps model, with a senior architect supervising an AI-assisted build.itecs.ai/pricing
Writes tests: PartlyAutomated tests and regression checks are set up as part of the release pipeline ITECS builds, but nothing says an agent authors test suites for client code.We set up version control, CI/CD, secrets management, staging validation, automated tests, deployment approvals, retrieval health checks, and…itecs.ai/ai-devops
Whole-codebase work: PartlyAgents can be connected to codebases among other systems, but carrying a migration across a whole repository is never claimed on any page read.Secure connections to CRMs, helpdesks, file stores, databases, codebases, and APIsitecs.ai/custom-ai-agents
Ships it: YesITECS builds and operates the deployment path itself: production pipelines, approval gates, controlled rollout and documented rollback for AI apps and agents.Production deployment pipelines for AI apps, agents, and automationsitecs.ai/ai-devops
Done-for-you service: YesThis is a services business: ITECS engineers design, build, deploy, train staff on and then operate the agents for the client.ITECS builds custom AI agents that connect securely to your proprietary data, tools, folders, codebases, and business systems - with…itecs.ai/custom-ai-agents

How much it does unattended

Runs autonomously: PartlyAgents run multi-step work on their own but consequential actions are deliberately blocked behind an approval queue, so full unattended completion is designed out.Sensitive actions wait in an approval queue with the draft, the sources, and the proposed change. Staff approve, edit, or reject. Nothing is sent,…itecs.ai/custom-ai-agents
Multi-agent: YesOrchestrated multi-agent systems are a published build phase with their own price band, alongside end-to-end process redesign.An orchestrated multi-agent system or end-to-end redesign of a major workflow with integrations, monitoring, and rollout enablement.itecs.ai/pricing
Agent permissions: PartlyPermission scope is designed per build: agents call only approved tools, and out-of-policy or low-confidence steps are routed to a person instead of acted on.Permissions, data boundaries, confidence thresholds, and DLP rules are checked before the agent acts. Out-of-policy or low-confidence steps route to…itecs.ai/custom-ai-agents

Which models it runs on

Claude: YesClaude is named as a build target for both enterprise agents and desktop personal agents, with a dedicated Claude Cowork rollout service.Claude, ChatGPT, Gemini, Copilot, Codex, CLI, and open-source agent workflowsitecs.ai/custom-ai-agents
GPT: YesChatGPT, Codex, the OpenAI API and Azure OpenAI all appear as platforms ITECS builds workflows on when they fit the use case.We build custom workflows for AI applications, APIs, CLIs, project folders, and agent frameworks, including Claude, Codex, ChatGPT, Gemini, Microsoft…itecs.ai/custom-ai-agents
Open models: PartlyOpen-source models and open-source agent workflows are listed in scope, but no specific open-weight model or hosting arrangement is named anywhere read.Vendor-neutral architecture - we build with the right mix of ChatGPT, Claude, Gemini, Copilot, Codex, APIs, open-source models, and business…itecs.ai/custom-ai-agents
Model choice: YesITECS states it is not tied to one AI vendor and selects the model and platform per use case, so the buyer is not locked to one engine.Yes. ITECS is not tied to one AI vendor.itecs.ai/custom-ai-agents

Where you use it

On the command line: YesCLI workflows and CLI automations are an explicit deliverable, including scoped runtime secret injection through a CLI secrets tool.Claude projects, Codex workflows, CLI automations, RAG agents, and external-system integrations are all in scopeitecs.ai/custom-ai-agents
In your pipeline: PartlyITECS builds CI/CD, release gates and rollback for the client's AI stack, but this is pipeline work it operates, not an agent that comments on pull requests.ITECS operationalizes AI systems for Dallas businesses with secure CI/CD, model and prompt version control, RAG pipeline monitoring, cost controls,…itecs.ai/ai-devops
In a browser: PartlyEnterprise builds include a hosted review portal where staff approve actions, but ITECS publishes no general web application a buyer logs into.Human-in-the-loop review portal before any actionitecs.ai

Whose machine it runs on

Self-hosted: PartlyPersonal agents are built and run on the employee's own machine and local agents are delivered inside the client's workspace, while enterprise builds sit in a governed hosted environment.Designed and executed inside Claude Cowork, Codex, and Microsoft Copilot on the individual's own machine.itecs.ai

What it costs to run

How it meters: YesBilling is per engagement and per running agent, not per seat or token: fixed package fees, a required discovery fee credited to the build, then a monthly fee per production agent.1 production agent: $2,500/mo after launch; production operations are separate from the build feeitecs.ai/custom-ai-agents
Free tier: PartlyThere is no free product tier; the only no-cost step is a scoping intake that leads to a paid engagement.Start the No-Cost Intakeitecs.ai
Bring your own key: PartlyBuilds run on the client's own licences and workspaces - Copilot, Claude and ChatGPT seats the employees already hold - rather than on ITECS-resold inference.ITECS builds and delivers one local agent inside the client's workspace, ready for the sponsoring employee to use.itecs.ai/pricing

Buying it for a team

A company can buy it: YesEvery package is sold to an organisation, with leadership deliverables, tenant reviews and user-band subscriptions aimed at 50-300 employee companies.Executive-ready deliverable suitable for board or leadership reviewitecs.ai/pricing
Seat model: PartlyThe ongoing Managed Intelligence subscription is priced in user bands rather than per seat, and build work is priced per workflow with no seat count at all.MIS Core Up to 10 users $1,950/mo 4 optimization hoursitecs.ai/pricing
Pooled budget: YesPrepaid hour blocks are explicitly a shared pool any employee can draw against, with a monthly usage statement to one named contact.Any employee may draw against the blockitecs.ai/pricing
Admin controls: YesConstraining what AI and agents may do is a priced deliverable: acceptable-use policy, governance framework, data rules and human review paths.ITECS defines approved use cases, data rules, human review paths, and operating ownership before AI spreads.itecs.ai
Audit log: YesAudit logging is part of the standard build pattern, and release changes to prompts, models and retrieval are versioned with approval records.Audit-ready releases - prompt, model, retrieval, and code changes are versioned with deploy history and approval recordsitecs.ai/ai-devops

What happens to your code

Opt out of training: PartlyITECS scopes private or enterprise configurations to the chosen vendors' contractual data terms, so the protection comes from those terms rather than an ITECS guarantee.ITECS scopes private or enterprise configurations, permissions, DLP policies, credential isolation, audit logs, and training to the selected vendors'…itecs.ai/custom-ai-agents
Getting out: PartlyBuild artefacts are archived in an ITECS client vault and re-wired for a successor rather than rebuilt, but purchased capacity does expire: sessions at 90 days, retainer hours at 12 months, and MIS hours do not roll over.Sessions expire 90 days from purchase.itecs.ai/pricing